title
stringlengths 7
246
| abstract
stringlengths 6
3.31k
|
|---|---|
Intel nGraph: An Intermediate Representation, Compiler, and Executor for
Deep Learning | The Deep Learning (DL) community sees many novel topologies published each
year. Achieving high performance on each new topology remains challenging, as
each requires some level of manual effort. This issue is compounded by the
proliferation of frameworks and hardware platforms. The current approach, which
we call "direct optimization", requires deep changes within each framework to
improve the training performance for each hardware backend (CPUs, GPUs, FPGAs,
ASICs) and requires $\mathcal{O}(fp)$ effort; where $f$ is the number of
frameworks and $p$ is the number of platforms. While optimized kernels for
deep-learning primitives are provided via libraries like Intel Math Kernel
Library for Deep Neural Networks (MKL-DNN), there are several compiler-inspired
ways in which performance can be further optimized. Building on our experience
creating neon (a fast deep learning library on GPUs), we developed Intel
nGraph, a soon to be open-sourced C++ library to simplify the realization of
optimized deep learning performance across frameworks and hardware platforms.
Initially-supported frameworks include TensorFlow, MXNet, and Intel neon
framework. Initial backends are Intel Architecture CPUs (CPU), the Intel(R)
Nervana Neural Network Processor(R) (NNP), and NVIDIA GPUs. Currently supported
compiler optimizations include efficient memory management and data layout
abstraction. In this paper, we describe our overall architecture and its core
components. In the future, we envision extending nGraph API support to a wider
range of frameworks, hardware (including FPGAs and ASICs), and compiler
optimizations (training versus inference optimizations, multi-node and
multi-device scaling via efficient sub-graph partitioning, and HW-specific
compounding of operations).
|
Generalizable Data-free Objective for Crafting Universal Adversarial
Perturbations | Machine learning models are susceptible to adversarial perturbations: small
changes to input that can cause large changes in output. It is also
demonstrated that there exist input-agnostic perturbations, called universal
adversarial perturbations, which can change the inference of target model on
most of the data samples. However, existing methods to craft universal
perturbations are (i) task specific, (ii) require samples from the training
data distribution, and (iii) perform complex optimizations. Additionally,
because of the data dependence, fooling ability of the crafted perturbations is
proportional to the available training data. In this paper, we present a novel,
generalizable and data-free approaches for crafting universal adversarial
perturbations. Independent of the underlying task, our objective achieves
fooling via corrupting the extracted features at multiple layers. Therefore,
the proposed objective is generalizable to craft image-agnostic perturbations
across multiple vision tasks such as object recognition, semantic segmentation,
and depth estimation. In the practical setting of black-box attack scenario
(when the attacker does not have access to the target model and it's training
data), we show that our objective outperforms the data dependent objectives to
fool the learned models. Further, via exploiting simple priors related to the
data distribution, our objective remarkably boosts the fooling ability of the
crafted perturbations. Significant fooling rates achieved by our objective
emphasize that the current deep learning models are now at an increased risk,
since our objective generalizes across multiple tasks without the requirement
of training data for crafting the perturbations. To encourage reproducible
research, we have released the codes for our proposed algorithm.
|
Learning Symmetric and Low-energy Locomotion | Learning locomotion skills is a challenging problem. To generate realistic
and smooth locomotion, existing methods use motion capture, finite state
machines or morphology-specific knowledge to guide the motion generation
algorithms. Deep reinforcement learning (DRL) is a promising approach for the
automatic creation of locomotion control. Indeed, a standard benchmark for DRL
is to automatically create a running controller for a biped character from a
simple reward function. Although several different DRL algorithms can
successfully create a running controller, the resulting motions usually look
nothing like a real runner. This paper takes a minimalist learning approach to
the locomotion problem, without the use of motion examples, finite state
machines, or morphology-specific knowledge. We introduce two modifications to
the DRL approach that, when used together, produce locomotion behaviors that
are symmetric, low-energy, and much closer to that of a real person. First, we
introduce a new term to the loss function (not the reward function) that
encourages symmetric actions. Second, we introduce a new curriculum learning
method that provides modulated physical assistance to help the character with
left/right balance and forward movement. The algorithm automatically computes
appropriate assistance to the character and gradually relaxes this assistance,
so that eventually the character learns to move entirely without help. Because
our method does not make use of motion capture data, it can be applied to a
variety of character morphologies. We demonstrate locomotion controllers for
the lower half of a biped, a full humanoid, a quadruped, and a hexapod. Our
results show that learned policies are able to produce symmetric, low-energy
gaits. In addition, speed-appropriate gait patterns emerge without any guidance
from motion examples or contact planning.
|
Adaptive Recurrent Neural Network Based on Mixture Layer | Although Recurrent Neural Network (RNN) has been a powerful tool for modeling
sequential data, its performance is inadequate when processing sequences with
multiple patterns. In this paper, we address this challenge by introducing a
novel mixture layer and constructing an adaptive RNN. The mixture layer
augmented RNN (termed as M-RNN) partitions patterns in training sequences into
several clusters and stores the principle patterns as prototype vectors of
components in a mixture model. By leveraging the mixture layer, the proposed
method can adaptively update states according to the similarities between
encoded inputs and prototype vectors, leading to a stronger capacity in
assimilating sequences with multiple patterns. Moreover, our approach can be
further extended by taking advantage of prior knowledge about data. Experiments
on both synthetic and real datasets demonstrate the effectiveness of the
proposed method.
|
Logically-Constrained Reinforcement Learning | We present the first model-free Reinforcement Learning (RL) algorithm to
synthesise policies for an unknown Markov Decision Process (MDP), such that a
linear time property is satisfied. The given temporal property is converted
into a Limit Deterministic Buchi Automaton (LDBA) and a robust reward function
is defined over the state-action pairs of the MDP according to the resulting
LDBA. With this reward function, the policy synthesis procedure is
"constrained" by the given specification. These constraints guide the MDP
exploration so as to minimize the solution time by only considering the portion
of the MDP that is relevant to satisfaction of the LTL property. This improves
performance and scalability of the proposed method by avoiding an exhaustive
update over the whole state space while the efficiency of standard methods such
as dynamic programming is hindered by excessive memory requirements, caused by
the need to store a full-model in memory. Additionally, we show that the RL
procedure sets up a local value iteration method to efficiently calculate the
maximum probability of satisfying the given property, at any given state of the
MDP. We prove that our algorithm is guaranteed to find a policy whose traces
probabilistically satisfy the LTL property if such a policy exists, and
additionally we show that our method produces reasonable control policies even
when the LTL property cannot be satisfied. The performance of the algorithm is
evaluated via a set of numerical examples. We observe an improvement of one
order of magnitude in the number of iterations required for the synthesis
compared to existing approaches.
|
Incremental Eigenpair Computation for Graph Laplacian Matrices: Theory
and Applications | The smallest eigenvalues and the associated eigenvectors (i.e., eigenpairs)
of a graph Laplacian matrix have been widely used in spectral clustering and
community detection. However, in real-life applications the number of clusters
or communities (say, $K$) is generally unknown a-priori. Consequently, the
majority of the existing methods either choose $K$ heuristically or they repeat
the clustering method with different choices of $K$ and accept the best
clustering result. The first option, more often, yields suboptimal result,
while the second option is computationally expensive. In this work, we propose
an incremental method for constructing the eigenspectrum of the graph Laplacian
matrix. This method leverages the eigenstructure of graph Laplacian matrix to
obtain the $K$-th smallest eigenpair of the Laplacian matrix given a collection
of all previously computed $K-1$ smallest eigenpairs. Our proposed method
adapts the Laplacian matrix such that the batch eigenvalue decomposition
problem transforms into an efficient sequential leading eigenpair computation
problem. As a practical application, we consider user-guided spectral
clustering. Specifically, we demonstrate that users can utilize the proposed
incremental method for effective eigenpair computation and for determining the
desired number of clusters based on multiple clustering metrics.
|
Active Neural Localization | Localization is the problem of estimating the location of an autonomous agent
from an observation and a map of the environment. Traditional methods of
localization, which filter the belief based on the observations, are
sub-optimal in the number of steps required, as they do not decide the actions
taken by the agent. We propose "Active Neural Localizer", a fully
differentiable neural network that learns to localize accurately and
efficiently. The proposed model incorporates ideas of traditional
filtering-based localization methods, by using a structured belief of the state
with multiplicative interactions to propagate belief, and combines it with a
policy model to localize accurately while minimizing the number of steps
required for localization. Active Neural Localizer is trained end-to-end with
reinforcement learning. We use a variety of simulation environments for our
experiments which include random 2D mazes, random mazes in the Doom game engine
and a photo-realistic environment in the Unreal game engine. The results on the
2D environments show the effectiveness of the learned policy in an idealistic
setting while results on the 3D environments demonstrate the model's capability
of learning the policy and perceptual model jointly from raw-pixel based RGB
observations. We also show that a model trained on random textures in the Doom
environment generalizes well to a photo-realistic office space environment in
the Unreal engine.
|
DKN: Deep Knowledge-Aware Network for News Recommendation | Online news recommender systems aim to address the information explosion of
news and make personalized recommendation for users. In general, news language
is highly condensed, full of knowledge entities and common sense. However,
existing methods are unaware of such external knowledge and cannot fully
discover latent knowledge-level connections among news. The recommended results
for a user are consequently limited to simple patterns and cannot be extended
reasonably. Moreover, news recommendation also faces the challenges of high
time-sensitivity of news and dynamic diversity of users' interests. To solve
the above problems, in this paper, we propose a deep knowledge-aware network
(DKN) that incorporates knowledge graph representation into news
recommendation. DKN is a content-based deep recommendation framework for
click-through rate prediction. The key component of DKN is a multi-channel and
word-entity-aligned knowledge-aware convolutional neural network (KCNN) that
fuses semantic-level and knowledge-level representations of news. KCNN treats
words and entities as multiple channels, and explicitly keeps their alignment
relationship during convolution. In addition, to address users' diverse
interests, we also design an attention module in DKN to dynamically aggregate a
user's history with respect to current candidate news. Through extensive
experiments on a real online news platform, we demonstrate that DKN achieves
substantial gains over state-of-the-art deep recommendation models. We also
validate the efficacy of the usage of knowledge in DKN.
|
NDDR-CNN: Layerwise Feature Fusing in Multi-Task CNNs by Neural
Discriminative Dimensionality Reduction | In this paper, we propose a novel Convolutional Neural Network (CNN)
structure for general-purpose multi-task learning (MTL), which enables
automatic feature fusing at every layer from different tasks. This is in
contrast with the most widely used MTL CNN structures which empirically or
heuristically share features on some specific layers (e.g., share all the
features except the last convolutional layer). The proposed layerwise feature
fusing scheme is formulated by combining existing CNN components in a novel
way, with clear mathematical interpretability as discriminative dimensionality
reduction, which is referred to as Neural Discriminative Dimensionality
Reduction (NDDR). Specifically, we first concatenate features with the same
spatial resolution from different tasks according to their channel dimension.
Then, we show that the discriminative dimensionality reduction can be fulfilled
by 1x1 Convolution, Batch Normalization, and Weight Decay in one CNN. The use
of existing CNN components ensures the end-to-end training and the
extensibility of the proposed NDDR layer to various state-of-the-art CNN
architectures in a "plug-and-play" manner. The detailed ablation analysis shows
that the proposed NDDR layer is easy to train and also robust to different
hyperparameters. Experiments on different task sets with various base network
architectures demonstrate the promising performance and desirable
generalizability of our proposed method. The code of our paper is available at
https://github.com/ethanygao/NDDR-CNN.
|
DeepWriting: Making Digital Ink Editable via Deep Generative Modeling | Digital ink promises to combine the flexibility and aesthetics of handwriting
and the ability to process, search and edit digital text. Character recognition
converts handwritten text into a digital representation, albeit at the cost of
losing personalized appearance due to the technical difficulties of separating
the interwoven components of content and style. In this paper, we propose a
novel generative neural network architecture that is capable of disentangling
style from content and thus making digital ink editable. Our model can
synthesize arbitrary text, while giving users control over the visual
appearance (style). For example, allowing for style transfer without changing
the content, editing of digital ink at the word level and other application
scenarios such as spell-checking and correction of handwritten text. We
furthermore contribute a new dataset of handwritten text with fine-grained
annotations at the character level and report results from an initial user
evaluation.
|
Data-Driven Impulse Response Regularization via Deep Learning | We consider the problem of impulse response estimation of stable linear
single-input single-output systems. It is a well-studied problem where flexible
non-parametric models recently offered a leap in performance compared to the
classical finite-dimensional model structures. Inspired by this development and
the success of deep learning we propose a new flexible data-driven model. Our
experiments indicate that the new model is capable of exploiting even more of
the hidden patterns that are present in the input-output data as compared to
the non-parametric models.
|
CommanderSong: A Systematic Approach for Practical Adversarial Voice
Recognition | The popularity of ASR (automatic speech recognition) systems, like Google
Voice, Cortana, brings in security concerns, as demonstrated by recent attacks.
The impacts of such threats, however, are less clear, since they are either
less stealthy (producing noise-like voice commands) or requiring the physical
presence of an attack device (using ultrasound). In this paper, we demonstrate
that not only are more practical and surreptitious attacks feasible but they
can even be automatically constructed. Specifically, we find that the voice
commands can be stealthily embedded into songs, which, when played, can
effectively control the target system through ASR without being noticed. For
this purpose, we developed novel techniques that address a key technical
challenge: integrating the commands into a song in a way that can be
effectively recognized by ASR through the air, in the presence of background
noise, while not being detected by a human listener. Our research shows that
this can be done automatically against real world ASR applications. We also
demonstrate that such CommanderSongs can be spread through Internet (e.g.,
YouTube) and radio, potentially affecting millions of ASR users. We further
present a new mitigation technique that controls this threat.
|
Deep Learning in Pharmacogenomics: From Gene Regulation to Patient
Stratification | This Perspective provides examples of current and future applications of deep
learning in pharmacogenomics, including: (1) identification of novel regulatory
variants located in noncoding domains and their function as applied to
pharmacoepigenomics; (2) patient stratification from medical records; and (3)
prediction of drugs, targets, and their interactions. Deep learning
encapsulates a family of machine learning algorithms that over the last decade
has transformed many important subfields of artificial intelligence (AI) and
has demonstrated breakthrough performance improvements on a wide range of tasks
in biomedicine. We anticipate that in the future deep learning will be widely
used to predict personalized drug response and optimize medication selection
and dosing, using knowledge extracted from large and complex molecular,
epidemiological, clinical, and demographic datasets.
|
Effective Building Block Design for Deep Convolutional Neural Networks
using Search | Deep learning has shown promising results on many machine learning tasks but
DL models are often complex networks with large number of neurons and layers,
and recently, complex layer structures known as building blocks. Finding the
best deep model requires a combination of finding both the right architecture
and the correct set of parameters appropriate for that architecture. In
addition, this complexity (in terms of layer types, number of neurons, and
number of layers) also present problems with generalization since larger
networks are easier to overfit to the data. In this paper, we propose a search
framework for finding effective architectural building blocks for convolutional
neural networks (CNN). Our approach is much faster at finding models that are
close to state-of-the-art in performance. In addition, the models discovered by
our approach are also smaller than models discovered by similar techniques. We
achieve these twin advantages by designing our search space in such a way that
it searches over a reduced set of state-of-the-art building blocks for CNNs
including residual block, inception block, inception-residual block, ResNeXt
block and many others. We apply this technique to generate models for multiple
image datasets and show that these models achieve performance comparable to
state-of-the-art (and even surpassing the state-of-the-art in one case). We
also show that learned models are transferable between datasets.
|
JointDNN: An Efficient Training and Inference Engine for Intelligent
Mobile Cloud Computing Services | Deep learning models are being deployed in many mobile intelligent
applications. End-side services, such as intelligent personal assistants,
autonomous cars, and smart home services often employ either simple local
models on the mobile or complex remote models on the cloud. However, recent
studies have shown that partitioning the DNN computations between the mobile
and cloud can increase the latency and energy efficiencies. In this paper, we
propose an efficient, adaptive, and practical engine, JointDNN, for
collaborative computation between a mobile device and cloud for DNNs in both
inference and training phase. JointDNN not only provides an energy and
performance efficient method of querying DNNs for the mobile side but also
benefits the cloud server by reducing the amount of its workload and
communications compared to the cloud-only approach. Given the DNN architecture,
we investigate the efficiency of processing some layers on the mobile device
and some layers on the cloud server. We provide optimization formulations at
layer granularity for forward- and backward-propagations in DNNs, which can
adapt to mobile battery limitations and cloud server load constraints and
quality of service. JointDNN achieves up to 18 and 32 times reductions on the
latency and mobile energy consumption of querying DNNs compared to the
status-quo approaches, respectively.
|
Quantization Error as a Metric for Dynamic Precision Scaling in Neural
Net Training | Recent work has explored reduced numerical precision for parameters,
activations, and gradients during neural network training as a way to reduce
the computational cost of training (Na & Mukhopadhyay, 2016) (Courbariaux et
al., 2014). We present a novel dynamic precision scaling (DPS) scheme. Using
stochastic fixed-point rounding, a quantization-error based scaling scheme, and
dynamic bit-widths during training, we achieve 98.8% test accuracy on the MNIST
dataset using an average bit-width of just 16 bits for weights and 14 bits for
activations, compared to the standard 32-bit floating point values used in deep
learning frameworks.
|
Considerations When Learning Additive Explanations for Black-Box Models | Many methods to explain black-box models, whether local or global, are
additive. In this paper, we study global additive explanations for non-additive
models, focusing on four explanation methods: partial dependence, Shapley
explanations adapted to a global setting, distilled additive explanations, and
gradient-based explanations. We show that different explanation methods
characterize non-additive components in a black-box model's prediction function
in different ways. We use the concepts of main and total effects to anchor
additive explanations, and quantitatively evaluate additive and non-additive
explanations. Even though distilled explanations are generally the most
accurate additive explanations, non-additive explanations such as tree
explanations that explicitly model non-additive components tend to be even more
accurate. Despite this, our user study showed that machine learning
practitioners were better able to leverage additive explanations for various
tasks. These considerations should be taken into account when considering which
explanation to trust and use to explain black-box models.
|
Neural Algebra of Classifiers | The world is fundamentally compositional, so it is natural to think of visual
recognition as the recognition of basic visually primitives that are composed
according to well-defined rules. This strategy allows us to recognize unseen
complex concepts from simple visual primitives. However, the current trend in
visual recognition follows a data greedy approach where huge amounts of data
are required to learn models for any desired visual concept. In this paper, we
build on the compositionality principle and develop an "algebra" to compose
classifiers for complex visual concepts. To this end, we learn neural network
modules to perform boolean algebra operations on simple visual classifiers.
Since these modules form a complete functional set, a classifier for any
complex visual concept defined as a boolean expression of primitives can be
obtained by recursively applying the learned modules, even if we do not have a
single training sample. As our experiments show, using such a framework, we can
compose classifiers for complex visual concepts outperforming standard
baselines on two well-known visual recognition benchmarks. Finally, we present
a qualitative analysis of our method and its properties.
|
PDNet: Semantic Segmentation integrated with a Primal-Dual Network for
Document binarization | Binarization of digital documents is the task of classifying each pixel in an
image of the document as belonging to the background (parchment/paper) or
foreground (text/ink). Historical documents are often subjected to
degradations, that make the task challenging. In the current work a deep neural
network architecture is proposed that combines a fully convolutional network
with an unrolled primal-dual network that can be trained end-to-end to achieve
state of the art binarization on four out of seven datasets. Document
binarization is formulated as an energy minimization problem. A fully
convolutional neural network is trained for semantic segmentation of pixels
that provides labeling cost associated with each pixel. This cost estimate is
refined along the edges to compensate for any over or under estimation of the
foreground class using a primal-dual approach. We provide necessary overview on
proximal operator that facilitates theoretical underpinning required to train a
primal-dual network using a gradient descent algorithm. Numerical instabilities
encountered due to the recurrent nature of primal-dual approach are handled. We
provide experimental results on document binarization competition dataset along
with network changes and hyperparameter tuning required for stability and
performance of the network. The network when pre-trained on synthetic dataset
performs better as per the competition metrics.
|
Improving Bi-directional Generation between Different Modalities with
Variational Autoencoders | We investigate deep generative models that can exchange multiple modalities
bi-directionally, e.g., generating images from corresponding texts and vice
versa. A major approach to achieve this objective is to train a model that
integrates all the information of different modalities into a joint
representation and then to generate one modality from the corresponding other
modality via this joint representation. We simply applied this approach to
variational autoencoders (VAEs), which we call a joint multimodal variational
autoencoder (JMVAE). However, we found that when this model attempts to
generate a large dimensional modality missing at the input, the joint
representation collapses and this modality cannot be generated successfully.
Furthermore, we confirmed that this difficulty cannot be resolved even using a
known solution. Therefore, in this study, we propose two models to prevent this
difficulty: JMVAE-kl and JMVAE-h. Results of our experiments demonstrate that
these methods can prevent the difficulty above and that they generate
modalities bi-directionally with equal or higher likelihood than conventional
VAE methods, which generate in only one direction. Moreover, we confirm that
these methods can obtain the joint representation appropriately, so that they
can generate various variations of modality by moving over the joint
representation or changing the value of another modality.
|
Classification of sparsely labeled spatio-temporal data through
semi-supervised adversarial learning | In recent years, Generative Adversarial Networks (GAN) have emerged as a
powerful method for learning the mapping from noisy latent spaces to realistic
data samples in high-dimensional space. So far, the development and application
of GANs have been predominantly focused on spatial data such as images. In this
project, we aim at modeling of spatio-temporal sensor data instead, i.e.
dynamic data over time. The main goal is to encode temporal data into a global
and low-dimensional latent vector that captures the dynamics of the
spatio-temporal signal. To this end, we incorporate auto-regressive RNNs,
Wasserstein GAN loss, spectral norm weight constraints and a semi-supervised
learning scheme into InfoGAN, a method for retrieval of meaningful latents in
adversarial learning. To demonstrate the modeling capability of our method, we
encode full-body skeletal human motion from a large dataset representing 60
classes of daily activities, recorded in a multi-Kinect setup. Initial results
indicate competitive classification performance of the learned latent
representations, compared to direct CNN/RNN inference. In future work, we plan
to apply this method on a related problem in the medical domain, i.e. on
recovery of meaningful latents in gait analysis of patients with vertigo and
balance disorders.
|
Multivariate normal mixture modeling, clustering and classification with
the rebmix package | The rebmix package provides R functions for random univariate and
multivariate finite mixture model generation, estimation, clustering and
classification. The paper is focused on multivariate normal mixture models with
unrestricted variance-covariance matrices. The objective is to show how to
generate datasets for a known number of components, numbers of observations and
component parameters, how to estimate the number of components, component
weights and component parameters and how to predict cluster and class
membership based upon a model trained by the REBMIX algorithm. The accompanying
plotting, bootstrapping and other features of the package are dealt with, too.
For demonstration purpose a multivariate normal dataset with unrestricted
variance-covariance matrices is studied.
|
Correlated Components Analysis - Extracting Reliable Dimensions in
Multivariate Data | How does one find dimensions in multivariate data that are reliably expressed
across repetitions? For example, in a brain imaging study one may want to
identify combinations of neural signals that are reliably expressed across
multiple trials or subjects. For a behavioral assessment with multiple ratings,
one may want to identify an aggregate score that is reliably reproduced across
raters. Correlated Components Analysis (CorrCA) addresses this problem by
identifying components that are maximally correlated between repetitions (e.g.
trials, subjects, raters). Here we formalize this as the maximization of the
ratio of between-repetition to within-repetition covariance. We show that this
criterion maximizes repeat-reliability, defined as mean over variance across
repeats, and that it leads to CorrCA or to multi-set Canonical Correlation
Analysis, depending on the constraints. Surprisingly, we also find that CorrCA
is equivalent to Linear Discriminant Analysis for zero-mean signals, which
provides an unexpected link between classic concepts of multivariate analysis.
We present an exact parametric test of statistical significance based on the
F-statistic for normally distributed independent samples, and present and
validate shuffle statistics for the case of dependent samples. Regularization
and extension to non-linear mappings using kernels are also presented. The
algorithms are demonstrated on a series of data analysis applications, and we
provide all code and data required to reproduce the results.
|
Recasting Gradient-Based Meta-Learning as Hierarchical Bayes | Meta-learning allows an intelligent agent to leverage prior learning episodes
as a basis for quickly improving performance on a novel task. Bayesian
hierarchical modeling provides a theoretical framework for formalizing
meta-learning as inference for a set of parameters that are shared across
tasks. Here, we reformulate the model-agnostic meta-learning algorithm (MAML)
of Finn et al. (2017) as a method for probabilistic inference in a hierarchical
Bayesian model. In contrast to prior methods for meta-learning via hierarchical
Bayes, MAML is naturally applicable to complex function approximators through
its use of a scalable gradient descent procedure for posterior inference.
Furthermore, the identification of MAML as hierarchical Bayes provides a way to
understand the algorithm's operation as a meta-learning procedure, as well as
an opportunity to make use of computational strategies for efficient inference.
We use this opportunity to propose an improvement to the MAML algorithm that
makes use of techniques from approximate inference and curvature estimation.
|
Object category learning and retrieval with weak supervision | We consider the problem of retrieving objects from image data and learning to
classify them into meaningful semantic categories with minimal supervision. To
that end, we propose a fully differentiable unsupervised deep clustering
approach to learn semantic classes in an end-to-end fashion without individual
class labeling using only unlabeled object proposals. The key contributions of
our work are 1) a kmeans clustering objective where the clusters are learned as
parameters of the network and are represented as memory units, and 2)
simultaneously building a feature representation, or embedding, while learning
to cluster it. This approach shows promising results on two popular computer
vision datasets: on CIFAR10 for clustering objects, and on the more complex and
challenging Cityscapes dataset for semantically discovering classes which
visually correspond to cars, people, and bicycles. Currently, the only
supervision provided is segmentation objectness masks, but this method can be
extended to use an unsupervised objectness-based object generation mechanism
which will make the approach completely unsupervised.
|
Approximate Inference via Weighted Rademacher Complexity | Rademacher complexity is often used to characterize the learnability of a
hypothesis class and is known to be related to the class size. We leverage this
observation and introduce a new technique for estimating the size of an
arbitrary weighted set, defined as the sum of weights of all elements in the
set. Our technique provides upper and lower bounds on a novel generalization of
Rademacher complexity to the weighted setting in terms of the weighted set
size. This generalizes Massart's Lemma, a known upper bound on the Rademacher
complexity in terms of the unweighted set size. We show that the weighted
Rademacher complexity can be estimated by solving a randomly perturbed
optimization problem, allowing us to derive high-probability bounds on the size
of any weighted set. We apply our method to the problems of calculating the
partition function of an Ising model and computing propositional model counts
(#SAT). Our experiments demonstrate that we can produce tighter bounds than
competing methods in both the weighted and unweighted settings.
|
Algorithmic Linearly Constrained Gaussian Processes | We algorithmically construct multi-output Gaussian process priors which
satisfy linear differential equations. Our approach attempts to parametrize all
solutions of the equations using Gr\"obner bases. If successful, a push forward
Gaussian process along the paramerization is the desired prior. We consider
several examples from physics, geomathematics and control, among them the full
inhomogeneous system of Maxwell's equations. By bringing together stochastic
learning and computer algebra in a novel way, we combine noisy observations
with precise algebraic computations.
|
Less is more: sampling chemical space with active learning | The development of accurate and transferable machine learning (ML) potentials
for predicting molecular energetics is a challenging task. The process of data
generation to train such ML potentials is a task neither well understood nor
researched in detail. In this work, we present a fully automated approach for
the generation of datasets with the intent of training universal ML potentials.
It is based on the concept of active learning (AL) via Query by Committee
(QBC), which uses the disagreement between an ensemble of ML potentials to
infer the reliability of the ensemble's prediction. QBC allows the presented AL
algorithm to automatically sample regions of chemical space where the ML
potential fails to accurately predict the potential energy. AL improves the
overall fitness of ANAKIN-ME (ANI) deep learning potentials in rigorous test
cases by mitigating human biases in deciding what new training data to use. AL
also reduces the training set size to a fraction of the data required when
using naive random sampling techniques. To provide validation of our AL
approach we develop the COMP6 benchmark (publicly available on GitHub), which
contains a diverse set of organic molecules. Through the AL process, it is
shown that the AL-based potentials perform as well as the ANI-1 potential on
COMP6 with only 10% of the data, and vastly outperforms ANI-1 with 25% the
amount of data. Finally, we show that our proposed AL technique develops a
universal ANI potential (ANI-1x) that provides accurate energy and force
predictions on the entire COMP6 benchmark. This universal ML potential achieves
a level of accuracy on par with the best ML potentials for single molecule or
materials, while remaining applicable to the general class of organic molecules
comprised of the elements CHNO.
|
Document Image Classification with Intra-Domain Transfer Learning and
Stacked Generalization of Deep Convolutional Neural Networks | In this work, a region-based Deep Convolutional Neural Network framework is
proposed for document structure learning. The contribution of this work
involves efficient training of region based classifiers and effective
ensembling for document image classification. A primary level of `inter-domain'
transfer learning is used by exporting weights from a pre-trained VGG16
architecture on the ImageNet dataset to train a document classifier on whole
document images. Exploiting the nature of region based influence modelling, a
secondary level of `intra-domain' transfer learning is used for rapid training
of deep learning models for image segments. Finally, stacked generalization
based ensembling is utilized for combining the predictions of the base deep
neural network models. The proposed method achieves state-of-the-art accuracy
of 92.2% on the popular RVL-CDIP document image dataset, exceeding benchmarks
set by existing algorithms.
|
Stochastic Downsampling for Cost-Adjustable Inference and Improved
Regularization in Convolutional Networks | It is desirable to train convolutional networks (CNNs) to run more
efficiently during inference. In many cases however, the computational budget
that the system has for inference cannot be known beforehand during training,
or the inference budget is dependent on the changing real-time resource
availability. Thus, it is inadequate to train just inference-efficient CNNs,
whose inference costs are not adjustable and cannot adapt to varied inference
budgets. We propose a novel approach for cost-adjustable inference in CNNs -
Stochastic Downsampling Point (SDPoint). During training, SDPoint applies
feature map downsampling to a random point in the layer hierarchy, with a
random downsampling ratio. The different stochastic downsampling configurations
known as SDPoint instances (of the same model) have computational costs
different from each other, while being trained to minimize the same prediction
loss. Sharing network parameters across different instances provides
significant regularization boost. During inference, one may handpick a SDPoint
instance that best fits the inference budget. The effectiveness of SDPoint, as
both a cost-adjustable inference approach and a regularizer, is validated
through extensive experiments on image classification.
|
Certified Defenses against Adversarial Examples | While neural networks have achieved high accuracy on standard image
classification benchmarks, their accuracy drops to nearly zero in the presence
of small adversarial perturbations to test inputs. Defenses based on
regularization and adversarial training have been proposed, but often followed
by new, stronger attacks that defeat these defenses. Can we somehow end this
arms race? In this work, we study this problem for neural networks with one
hidden layer. We first propose a method based on a semidefinite relaxation that
outputs a certificate that for a given network and test input, no attack can
force the error to exceed a certain value. Second, as this certificate is
differentiable, we jointly optimize it with the network parameters, providing
an adaptive regularizer that encourages robustness against all attacks. On
MNIST, our approach produces a network and a certificate that no attack that
perturbs each pixel by at most \epsilon = 0.1 can cause more than 35% test
error.
|
On the Inter-relationships among Drift rate, Forgetting rate,
Bias/variance profile and Error | We propose two general and falsifiable hypotheses about expectations on
generalization error when learning in the context of concept drift. One posits
that as drift rate increases, the forgetting rate that minimizes generalization
error will also increase and vice versa. The other posits that as a learner's
forgetting rate increases, the bias/variance profile that minimizes
generalization error will have lower variance and vice versa. These hypotheses
lead to the concept of the sweet path, a path through the 3-d space of
alternative drift rates, forgetting rates and bias/variance profiles on which
generalization error will be minimized, such that slow drift is coupled with
low forgetting and low bias, while rapid drift is coupled with fast forgetting
and low variance. We present experiments that support the existence of such a
sweet path. We also demonstrate that simple learners that select appropriate
forgetting rates and bias/variance profiles are highly competitive with the
state-of-the-art in incremental learners for concept drift on real-world drift
problems.
|
Nonlinear Dimensionality Reduction on Graphs | In this era of data deluge, many signal processing and machine learning tasks
are faced with high-dimensional datasets, including images, videos, as well as
time series generated from social, commercial and brain network interactions.
Their efficient processing calls for dimensionality reduction techniques
capable of properly compressing the data while preserving task-related
characteristics, going beyond pairwise data correlations. The present paper
puts forth a nonlinear dimensionality reduction framework that accounts for
data lying on known graphs. The novel framework encompasses most of the
existing dimensionality reduction methods, but it is also capable of capturing
and preserving possibly nonlinear correlations that are ignored by linear
methods. Furthermore, it can take into account information from multiple
graphs. The proposed algorithms were tested on synthetic as well as real
datasets to corroborate their effectiveness.
|
Learning Combinations of Activation Functions | In the last decade, an active area of research has been devoted to design
novel activation functions that are able to help deep neural networks to
converge, obtaining better performance. The training procedure of these
architectures usually involves optimization of the weights of their layers
only, while non-linearities are generally pre-specified and their (possible)
parameters are usually considered as hyper-parameters to be tuned manually. In
this paper, we introduce two approaches to automatically learn different
combinations of base activation functions (such as the identity function, ReLU,
and tanh) during the training phase. We present a thorough comparison of our
novel approaches with well-known architectures (such as LeNet-5, AlexNet, and
ResNet-56) on three standard datasets (Fashion-MNIST, CIFAR-10, and
ILSVRC-2012), showing substantial improvements in the overall performance, such
as an increase in the top-1 accuracy for AlexNet on ILSVRC-2012 of 3.01
percentage points.
|
A notion of stability for k-means clustering | In this paper, we define and study a new notion of stability for the
$k$-means clustering scheme building upon the notion of quantization of a
probability measure. We connect this notion of stability to a geometric feature
of the underlying distribution of the data, named absolute margin condition,
inspired by recent works on the subject.
|
Improving Active Learning in Systematic Reviews | Systematic reviews are essential to summarizing the results of different
clinical and social science studies. The first step in a systematic review task
is to identify all the studies relevant to the review. The task of identifying
relevant studies for a given systematic review is usually performed manually,
and as a result, involves substantial amounts of expensive human resource.
Lately, there have been some attempts to reduce this manual effort using active
learning. In this work, we build upon some such existing techniques, and
validate by experimenting on a larger and comprehensive dataset than has been
attempted until now. Our experiments provide insights on the use of different
feature extraction models for different disciplines. More importantly, we
identify that a naive active learning based screening process is biased in
favour of selecting similar documents. We aimed to improve the performance of
the screening process using a novel active learning algorithm with success.
Additionally, we propose a mechanism to choose the best feature extraction
method for a given review.
|
Deep Learning Angiography (DLA): Three-dimensional C-arm Cone Beam CT
Angiography Using Deep Learning | Background and Purpose: Our purpose was to develop a deep learning
angiography (DLA) method to generate 3D cerebral angiograms from a single
contrast-enhanced acquisition.
Material and Methods: Under an approved IRB protocol 105 3D-DSA exams were
randomly selected from an internal database. All were acquired using a clinical
system (Axiom Artis zee, Siemens Healthineers) in conjunction with a standard
injection protocol. More than 150 million labeled voxels from 35 subjects were
used for training. A deep convolutional neural network was trained to classify
each image voxel into three tissue types (vasculature, bone and soft tissue).
The trained DLA model was then applied for tissue classification in a
validation cohort of 8 subjects and a final testing cohort consisting of the
remaining 62 subjects. The final vasculature tissue class was used to generate
the 3D-DLA images. To quantify the generalization error of the trained model,
accuracy, sensitivity, precision and F1-scores were calculated for vasculature
classification in relevant anatomy. The 3D-DLA and clinical 3D-DSA images were
subject to a qualitative assessment for the presence of inter-sweep motion
artifacts.
Results: Vasculature classification accuracy and 95% CI in the testing
dataset was 98.7% ([98.3, 99.1] %). No residual signal from osseous structures
was observed for all 3D-DLA testing cases except for small regions in the otic
capsule and nasal cavity compared to 37% (23/62) of the 3D-DSAs.
Conclusion: DLA accurately recreated the vascular anatomy of the 3D-DSA
reconstructions without mask. DLA reduced mis-registration artifacts induced by
inter-sweep motion. DLA reduces radiation exposure required to obtain
clinically useful 3D-DSA
|
Multichannel Sound Event Detection Using 3D Convolutional Neural
Networks for Learning Inter-channel Features | In this paper, we propose a stacked convolutional and recurrent neural
network (CRNN) with a 3D convolutional neural network (CNN) in the first layer
for the multichannel sound event detection (SED) task. The 3D CNN enables the
network to simultaneously learn the inter- and intra-channel features from the
input multichannel audio. In order to evaluate the proposed method,
multichannel audio datasets with different number of overlapping sound sources
are synthesized. Each of this dataset has a four-channel first-order Ambisonic,
binaural, and single-channel versions, on which the performance of SED using
the proposed method are compared to study the potential of SED using
multichannel audio. A similar study is also done with the binaural and
single-channel versions of the real-life recording TUT-SED 2017 development
dataset. The proposed method learns to recognize overlapping sound events from
multichannel features faster and performs better SED with a fewer number of
training epochs. The results show that on using multichannel Ambisonic audio in
place of single-channel audio we improve the overall F-score by 7.5%, overall
error rate by 10% and recognize 15.6% more sound events in time frames with
four overlapping sound sources.
|
21 Million Opportunities: A 19 Facility Investigation of Factors
Affecting Hand Hygiene Compliance via Linear Predictive Models | This large-scale study, consisting of 21.3 million hand hygiene opportunities
from 19 distinct facilities in 10 different states, uses linear predictive
models to expose factors that may affect hand hygiene compliance. We examine
the use of features such as temperature, relative humidity, influenza severity,
day/night shift, federal holidays and the presence of new medical residents in
predicting daily hand hygiene compliance; the investigation is undertaken using
both a "global" model to glean general trends, and facility-specific models to
elicit facility-specific insights. The results suggest that colder temperatures
and federal holidays have an adverse effect on hand hygiene compliance rates,
and that individual cultures and attitudes regarding hand hygiene exist among
facilities.
|
DeepLung: Deep 3D Dual Path Nets for Automated Pulmonary Nodule
Detection and Classification | In this work, we present a fully automated lung computed tomography (CT)
cancer diagnosis system, DeepLung. DeepLung consists of two components, nodule
detection (identifying the locations of candidate nodules) and classification
(classifying candidate nodules into benign or malignant). Considering the 3D
nature of lung CT data and the compactness of dual path networks (DPN), two
deep 3D DPN are designed for nodule detection and classification respectively.
Specifically, a 3D Faster Regions with Convolutional Neural Net (R-CNN) is
designed for nodule detection with 3D dual path blocks and a U-net-like
encoder-decoder structure to effectively learn nodule features. For nodule
classification, gradient boosting machine (GBM) with 3D dual path network
features is proposed. The nodule classification subnetwork was validated on a
public dataset from LIDC-IDRI, on which it achieved better performance than
state-of-the-art approaches and surpassed the performance of experienced
doctors based on image modality. Within the DeepLung system, candidate nodules
are detected first by the nodule detection subnetwork, and nodule diagnosis is
conducted by the classification subnetwork. Extensive experimental results
demonstrate that DeepLung has performance comparable to experienced doctors
both for the nodule-level and patient-level diagnosis on the LIDC-IDRI
dataset.\footnote{https://github.com/uci-cbcl/DeepLung.git}
|
Deep Learning Approach for Very Similar Objects Recognition Application
on Chihuahua and Muffin Problem | We address the problem to tackle the very similar objects like Chihuahua or
muffin problem to recognize at least in human vision level. Our regular deep
structured machine learning still does not solve it. We saw many times for
about year in our community the problem. Today we proposed the state-of-the-art
solution for it. Our approach is quite tricky to get the very high accuracy. We
propose the deep transfer learning method which could be tackled all this type
of problems not limited to just Chihuahua or muffin problem. It is the best
method to train with small data set not like require huge amount data.
|
Learning the Reward Function for a Misspecified Model | In model-based reinforcement learning it is typical to decouple the problems
of learning the dynamics model and learning the reward function. However, when
the dynamics model is flawed, it may generate erroneous states that would never
occur in the true environment. It is not clear a priori what value the reward
function should assign to such states. This paper presents a novel error bound
that accounts for the reward model's behavior in states sampled from the model.
This bound is used to extend the existing Hallucinated DAgger-MC algorithm,
which offers theoretical performance guarantees in deterministic MDPs that do
not assume a perfect model can be learned. Empirically, this approach to reward
learning can yield dramatic improvements in control performance when the
dynamics model is flawed.
|
Barrier-Certified Adaptive Reinforcement Learning with Applications to
Brushbot Navigation | This paper presents a safe learning framework that employs an adaptive model
learning algorithm together with barrier certificates for systems with possibly
nonstationary agent dynamics. To extract the dynamic structure of the model, we
use a sparse optimization technique. We use the learned model in combination
with control barrier certificates which constrain policies (feedback
controllers) in order to maintain safety, which refers to avoiding particular
undesirable regions of the state space. Under certain conditions, recovery of
safety in the sense of Lyapunov stability after violations of safety due to the
nonstationarity is guaranteed. In addition, we reformulate an action-value
function approximation to make any kernel-based nonlinear function estimation
method applicable to our adaptive learning framework. Lastly, solutions to the
barrier-certified policy optimization are guaranteed to be globally optimal,
ensuring the greedy policy improvement under mild conditions. The resulting
framework is validated via simulations of a quadrotor, which has previously
been used under stationarity assumptions in the safe learnings literature, and
is then tested on a real robot, the brushbot, whose dynamics is unknown, highly
complex and nonstationary.
|
Matrix Completion for Structured Observations | The need to predict or fill-in missing data, often referred to as matrix
completion, is a common challenge in today's data-driven world. Previous
strategies typically assume that no structural difference between observed and
missing entries exists. Unfortunately, this assumption is woefully unrealistic
in many applications. For example, in the classic Netflix challenge, in which
one hopes to predict user-movie ratings for unseen films, the fact that the
viewer has not watched a given movie may indicate a lack of interest in that
movie, thus suggesting a lower rating than otherwise expected. We propose
adjusting the standard nuclear norm minimization strategy for matrix completion
to account for such structural differences between observed and unobserved
entries by regularizing the values of the unobserved entries. We show that the
proposed method outperforms nuclear norm minimization in certain settings.
|
tempoGAN: A Temporally Coherent, Volumetric GAN for Super-resolution
Fluid Flow | We propose a temporally coherent generative model addressing the
super-resolution problem for fluid flows. Our work represents a first approach
to synthesize four-dimensional physics fields with neural networks. Based on a
conditional generative adversarial network that is designed for the inference
of three-dimensional volumetric data, our model generates consistent and
detailed results by using a novel temporal discriminator, in addition to the
commonly used spatial one. Our experiments show that the generator is able to
infer more realistic high-resolution details by using additional physical
quantities, such as low-resolution velocities or vorticities. Besides
improvements in the training process and in the generated outputs, these inputs
offer means for artistic control as well. We additionally employ a
physics-aware data augmentation step, which is crucial to avoid overfitting and
to reduce memory requirements. In this way, our network learns to generate
advected quantities with highly detailed, realistic, and temporally coherent
features. Our method works instantaneously, using only a single time-step of
low-resolution fluid data. We demonstrate the abilities of our method using a
variety of complex inputs and applications in two and three dimensions.
|
A Generalized Circuit for the Hamiltonian Dynamics Through the Truncated
Series | In this paper, we present a method for the Hamiltonian simulation in the
context of eigenvalue estimation problems which improves earlier results
dealing with Hamiltonian simulation through the truncated Taylor series. In
particular, we present a fixed-quantum circuit design for the simulation of the
Hamiltonian dynamics, $H(t)$, through the truncated Taylor series method
described by Berry et al. \cite{berry2015simulating}. The circuit is general
and can be used to simulate any given matrix in the phase estimation algorithm
by only changing the angle values of the quantum gates implementing the time
variable $t$ in the series. The circuit complexity depends on the number of
summation terms composing the Hamiltonian and requires $O(Ln)$ number of
quantum gates for the simulation of a molecular Hamiltonian. Here, $n$ is the
number of states of a spin orbital, and $L$ is the number of terms in the
molecular Hamiltonian and generally bounded by $O(n^4)$. We also discuss how to
use the circuit in adaptive processes and eigenvalue related problems along
with a slight modified version of the iterative phase estimation algorithm. In
addition, a simple divide and conquer method is presented for mapping a matrix
which are not given as sums of unitary matrices into the circuit. The
complexity of the circuit is directly related to the structure of the matrix
and can be bounded by $O(poly(n))$ for a matrix with $poly(n)-$sparsity.
|
Evaluating approaches for supervised semantic labeling | Relational data sources are still one of the most popular ways to store
enterprise or Web data, however, the issue with relational schema is the lack
of a well-defined semantic description. A common ontology provides a way to
represent the meaning of a relational schema and can facilitate the integration
of heterogeneous data sources within a domain. Semantic labeling is achieved by
mapping attributes from the data sources to the classes and properties in the
ontology. We formulate this problem as a multi-class classification problem
where previously labeled data sources are used to learn rules for labeling new
data sources. The majority of existing approaches for semantic labeling have
focused on data integration challenges such as naming conflicts and semantic
heterogeneity. In addition, machine learning approaches typically have issues
around class imbalance, lack of labeled instances and relative importance of
attributes. To address these issues, we develop a new machine learning model
with engineered features as well as two deep learning models which do not
require extensive feature engineering. We evaluate our new approaches with the
state-of-the-art.
|
Discrete Autoencoders for Sequence Models | Recurrent models for sequences have been recently successful at many tasks,
especially for language modeling and machine translation. Nevertheless, it
remains challenging to extract good representations from these models. For
instance, even though language has a clear hierarchical structure going from
characters through words to sentences, it is not apparent in current language
models. We propose to improve the representation in sequence models by
augmenting current approaches with an autoencoder that is forced to compress
the sequence through an intermediate discrete latent space. In order to
propagate gradients though this discrete representation we introduce an
improved semantic hashing technique. We show that this technique performs well
on a newly proposed quantitative efficiency measure. We also analyze latent
codes produced by the model showing how they correspond to words and phrases.
Finally, we present an application of the autoencoder-augmented model to
generating diverse translations.
|
The Intriguing Properties of Model Explanations | Linear approximations to the decision boundary of a complex model have become
one of the most popular tools for interpreting predictions. In this paper, we
study such linear explanations produced either post-hoc by a few recent methods
or generated along with predictions with contextual explanation networks
(CENs). We focus on two questions: (i) whether linear explanations are always
consistent or can be misleading, and (ii) when integrated into the prediction
process, whether and how explanations affect the performance of the model. Our
analysis sheds more light on certain properties of explanations produced by
different methods and suggests that learning models that explain and predict
jointly is often advantageous.
|
Personalized Survival Prediction with Contextual Explanation Networks | Accurate and transparent prediction of cancer survival times on the level of
individual patients can inform and improve patient care and treatment
practices. In this paper, we design a model that concurrently learns to
accurately predict patient-specific survival distributions and to explain its
predictions in terms of patient attributes such as clinical tests or
assessments. Our model is flexible and based on a recurrent network, can handle
various modalities of data including temporal measurements, and yet constructs
and uses simple explanations in the form of patient- and time-specific linear
regression. For analysis, we use two publicly available datasets and show that
our networks outperform a number of baselines in prediction while providing a
way to inspect the reasons behind each prediction.
|
Learning to Emulate an Expert Projective Cone Scheduler | Projective cone scheduling defines a large class of rate-stabilizing policies
for queueing models relevant to several applications. While there exists
considerable theory on the properties of projective cone schedulers, there is
little practical guidance on choosing the parameters that define them. In this
paper, we propose an algorithm for designing an automated projective cone
scheduling system based on observations of an expert projective cone scheduler.
We show that the estimated scheduling policy is able to emulate the expert in
the sense that the average loss realized by the learned policy will converge to
zero. Specifically, for a system with $n$ queues observed over a time horizon
$T$, the average loss for the algorithm is $O(\ln(T)\sqrt{\ln(n)/T})$. This
upper bound holds regardless of the statistical characteristics of the system.
The algorithm uses the multiplicative weights update method and can be applied
online so that additional observations of the expert scheduler can be used to
improve an existing estimate of the policy. This provides a data-driven method
for designing a scheduling policy based on observations of a human expert. We
demonstrate the efficacy of the algorithm with a simple numerical example and
discuss several extensions.
|
Robustness of classification ability of spiking neural networks | It is well-known that the robustness of artificial neural networks (ANNs) is
important for their wide ranges of applications. In this paper, we focus on the
robustness of the classification ability of a spiking neural network which
receives perturbed inputs. Actually, the perturbation is allowed to be
arbitrary styles. However, Gaussian perturbation and other regular ones have
been rarely investigated. For classification problems, the closer to the
desired point, the more perturbed points there are in the input space. In
addition, the perturbation may be periodic. Based on these facts, we only
consider sinusoidal and Gaussian perturbations in this paper. With the
SpikeProp algorithm, we perform extensive experiments on the classical XOR
problem and other three benchmark datasets. The numerical results show that
there is not significant reduction in the classification ability of the network
if the input signals are subject to sinusoidal and Gaussian perturbations.
|
ReNN: Rule-embedded Neural Networks | The artificial neural network shows powerful ability of inference, but it is
still criticized for lack of interpretability and prerequisite needs of big
dataset. This paper proposes the Rule-embedded Neural Network (ReNN) to
overcome the shortages. ReNN first makes local-based inferences to detect local
patterns, and then uses rules based on domain knowledge about the local
patterns to generate rule-modulated map. After that, ReNN makes global-based
inferences that synthesizes the local patterns and the rule-modulated map. To
solve the optimization problem caused by rules, we use a two-stage optimization
strategy to train the ReNN model. By introducing rules into ReNN, we can
strengthen traditional neural networks with long-term dependencies which are
difficult to learn with limited empirical dataset, thus improving inference
accuracy. The complexity of neural networks can be reduced since long-term
dependencies are not modeled with neural connections, and thus the amount of
data needed to optimize the neural networks can be reduced. Besides, inferences
from ReNN can be analyzed with both local patterns and rules, and thus have
better interpretability. In this paper, ReNN has been validated with a
time-series detection problem.
|
Accelerating recurrent neural network language model based online speech
recognition system | This paper presents methods to accelerate recurrent neural network based
language models (RNNLMs) for online speech recognition systems. Firstly, a
lossy compression of the past hidden layer outputs (history vector) with
caching is introduced in order to reduce the number of LM queries. Next, RNNLM
computations are deployed in a CPU-GPU hybrid manner, which computes each layer
of the model on a more advantageous platform. The added overhead by data
exchanges between CPU and GPU is compensated through a frame-wise batching
strategy. The performance of the proposed methods evaluated on LibriSpeech test
sets indicates that the reduction in history vector precision improves the
average recognition speed by 1.23 times with minimum degradation in accuracy.
On the other hand, the CPU-GPU hybrid parallelization enables RNNLM based
real-time recognition with a four times improvement in speed.
|
Fast Power system security analysis with Guided Dropout | We propose a new method to efficiently compute load-flows (the steady-state
of the power-grid for given productions, consumptions and grid topology),
substituting conventional simulators based on differential equation solvers. We
use a deep feed-forward neural network trained with load-flows precomputed by
simulation. Our architecture permits to train a network on so-called "n-1"
problems, in which load flows are evaluated for every possible line
disconnection, then generalize to "n-2" problems without retraining (a clear
advantage because of the combinatorial nature of the problem). To that end, we
developed a technique bearing similarity with "dropout", which we named "guided
dropout".
|
Binary Compressive Sensing via Smoothed $\ell_0$ Gradient Descent | We present a Compressive Sensing algorithm for reconstructing binary signals
from its linear measurements. The proposed algorithm minimizes a non-convex
cost function expressed as a weighted sum of smoothed $\ell_0$ norms which
takes into account the binariness of signals. We show that for binary signals
the proposed algorithm outperforms other existing algorithms in recovery rate
while requiring a short run time.
|
COBRA: A Fast and Simple Method for Active Clustering with Pairwise
Constraints | Clustering is inherently ill-posed: there often exist multiple valid
clusterings of a single dataset, and without any additional information a
clustering system has no way of knowing which clustering it should produce.
This motivates the use of constraints in clustering, as they allow users to
communicate their interests to the clustering system. Active constraint-based
clustering algorithms select the most useful constraints to query, aiming to
produce a good clustering using as few constraints as possible. We propose
COBRA, an active method that first over-clusters the data by running K-means
with a $K$ that is intended to be too large, and subsequently merges the
resulting small clusters into larger ones based on pairwise constraints. In its
merging step, COBRA is able to keep the number of pairwise queries low by
maximally exploiting constraint transitivity and entailment. We experimentally
show that COBRA outperforms the state of the art in terms of clustering quality
and runtime, without requiring the number of clusters in advance.
|
Cardiac Arrhythmia Detection from ECG Combining Convolutional and Long
Short-Term Memory Networks | Objectives: Atrial fibrillation (AF) is a common heart rhythm disorder
associated with deadly and debilitating consequences including heart failure,
stroke, poor mental health, reduced quality of life and death. Having an
automatic system that diagnoses various types of cardiac arrhythmias would
assist cardiologists to initiate appropriate preventive measures and to improve
the analysis of cardiac disease. To this end, this paper introduces a new
approach to detect and classify automatically cardiac arrhythmias in
electrocardiograms (ECG) recordings.
Methods: The proposed approach used a combination of Convolution Neural
Networks (CNNs) and a sequence of Long Short-Term Memory (LSTM) units, with
pooling, dropout and normalization techniques to improve their accuracy. The
network predicted a classification at every 18th input sample and we selected
the final prediction for classification. Results were cross-validated on the
Physionet Challenge 2017 training dataset, which contains 8,528 single lead ECG
recordings lasting from 9s to just over 60s.
Results: Using the proposed structure and no explicit feature selection,
10-fold stratified cross-validation gave an overall F-measure of 0.83.10-0.015
on the held-out test data (mean-standard deviation over all folds) and 0.80 on
the hidden dataset of the Challenge entry server.
|
Rigorous Restricted Isometry Property of Low-Dimensional Subspaces | Dimensionality reduction is in demand to reduce the complexity of solving
large-scale problems with data lying in latent low-dimensional structures in
machine learning and computer version. Motivated by such need, in this work we
study the Restricted Isometry Property (RIP) of Gaussian random projections for
low-dimensional subspaces in $\mathbb{R}^N$, and rigorously prove that the
projection Frobenius norm distance between any two subspaces spanned by the
projected data in $\mathbb{R}^n$ ($n<N$) remain almost the same as the distance
between the original subspaces with probability no less than $1 - {\rm
e}^{-\mathcal{O}(n)}$. Previously the well-known Johnson-Lindenstrauss (JL)
Lemma and RIP for sparse vectors have been the foundation of sparse signal
processing including Compressed Sensing. As an analogy to JL Lemma and RIP for
sparse vectors, this work allows the use of random projections to reduce the
ambient dimension with the theoretical guarantee that the distance between
subspaces after compression is well preserved.
|
Anomaly detection in wide area network mesh using two machine learning
anomaly detection algorithms | Anomaly detection is the practice of identifying items or events that do not
conform to an expected behavior or do not correlate with other items in a
dataset. It has previously been applied to areas such as intrusion detection,
system health monitoring, and fraud detection in credit card transactions. In
this paper, we describe a new method for detecting anomalous behavior over
network performance data, gathered by perfSONAR, using two machine learning
algorithms: Boosted Decision Trees (BDT) and Simple Feedforward Neural Network.
The effectiveness of each algorithm was evaluated and compared. Both have shown
sufficient performance and sensitivity.
|
An Incremental Path-Following Splitting Method for Linearly Constrained
Nonconvex Nonsmooth Programs | The stationary point of Problem 2 is NOT the stationary point of Problem 1.
We are sorry and we are working on fixing this error.
|
Links: A High-Dimensional Online Clustering Method | We present a novel algorithm, called Links, designed to perform online
clustering on unit vectors in a high-dimensional Euclidean space. The algorithm
is appropriate when it is necessary to cluster data efficiently as it streams
in, and is to be contrasted with traditional batch clustering algorithms that
have access to all data at once. For example, Links has been successfully
applied to embedding vectors generated from face images or voice recordings for
the purpose of recognizing people, thereby providing real-time identification
during video or audio capture.
|
Spherical CNNs | Convolutional Neural Networks (CNNs) have become the method of choice for
learning problems involving 2D planar images. However, a number of problems of
recent interest have created a demand for models that can analyze spherical
images. Examples include omnidirectional vision for drones, robots, and
autonomous cars, molecular regression problems, and global weather and climate
modelling. A naive application of convolutional networks to a planar projection
of the spherical signal is destined to fail, because the space-varying
distortions introduced by such a projection will make translational weight
sharing ineffective.
In this paper we introduce the building blocks for constructing spherical
CNNs. We propose a definition for the spherical cross-correlation that is both
expressive and rotation-equivariant. The spherical correlation satisfies a
generalized Fourier theorem, which allows us to compute it efficiently using a
generalized (non-commutative) Fast Fourier Transform (FFT) algorithm. We
demonstrate the computational efficiency, numerical accuracy, and effectiveness
of spherical CNNs applied to 3D model recognition and atomization energy
regression.
|
Sometimes You Want to Go Where Everybody Knows your Name | We introduce a new metric for measuring how well a model personalizes to a
user's specific preferences. We define personalization as a weighting between
performance on user specific data and performance on a more general global
dataset that represents many different users. This global term serves as a form
of regularization that forces us to not overfit to individual users who have
small amounts of data. In order to protect user privacy, we add the constraint
that we may not centralize or share user data. We also contribute a simple
experiment in which we simulate classifying sentiment for users with very
distinct vocabularies. This experiment functions as an example of the tension
between doing well globally on all users, and doing well on any specific
individual user. It also provides a concrete example of how to employ our new
metric to help reason about and resolve this tension. We hope this work can
help frame and ground future work into personalization.
|
DeepDTA: Deep Drug-Target Binding Affinity Prediction | The identification of novel drug-target (DT) interactions is a substantial
part of the drug discovery process. Most of the computational methods that have
been proposed to predict DT interactions have focused on binary classification,
where the goal is to determine whether a DT pair interacts or not. However,
protein-ligand interactions assume a continuum of binding strength values, also
called binding affinity and predicting this value still remains a challenge.
The increase in the affinity data available in DT knowledge-bases allows the
use of advanced learning techniques such as deep learning architectures in the
prediction of binding affinities. In this study, we propose a deep-learning
based model that uses only sequence information of both targets and drugs to
predict DT interaction binding affinities. The few studies that focus on DT
binding affinity prediction use either 3D structures of protein-ligand
complexes or 2D features of compounds. One novel approach used in this work is
the modeling of protein sequences and compound 1D representations with
convolutional neural networks (CNNs). The results show that the proposed deep
learning based model that uses the 1D representations of targets and drugs is
an effective approach for drug target binding affinity prediction. The model in
which high-level representations of a drug and a target are constructed via
CNNs achieved the best Concordance Index (CI) performance in one of our larger
benchmark data sets, outperforming the KronRLS algorithm and SimBoost, a
state-of-the-art method for DT binding affinity prediction.
|
Low-Rank Bandit Methods for High-Dimensional Dynamic Pricing | We consider dynamic pricing with many products under an evolving but
low-dimensional demand model. Assuming the temporal variation in
cross-elasticities exhibits low-rank structure based on fixed (latent) features
of the products, we show that the revenue maximization problem reduces to an
online bandit convex optimization with side information given by the observed
demands. We design dynamic pricing algorithms whose revenue approaches that of
the best fixed price vector in hindsight, at a rate that only depends on the
intrinsic rank of the demand model and not the number of products. Our approach
applies a bandit convex optimization algorithm in a projected low-dimensional
space spanned by the latent product features, while simultaneously learning
this span via online singular value decomposition of a carefully-crafted matrix
containing the observed demands.
|
FastGCN: Fast Learning with Graph Convolutional Networks via Importance
Sampling | The graph convolutional networks (GCN) recently proposed by Kipf and Welling
are an effective graph model for semi-supervised learning. This model, however,
was originally designed to be learned with the presence of both training and
test data. Moreover, the recursive neighborhood expansion across layers poses
time and memory challenges for training with large, dense graphs. To relax the
requirement of simultaneous availability of test data, we interpret graph
convolutions as integral transforms of embedding functions under probability
measures. Such an interpretation allows for the use of Monte Carlo approaches
to consistently estimate the integrals, which in turn leads to a batched
training scheme as we propose in this work---FastGCN. Enhanced with importance
sampling, FastGCN not only is efficient for training but also generalizes well
for inference. We show a comprehensive set of experiments to demonstrate its
effectiveness compared with GCN and related models. In particular, training is
orders of magnitude more efficient while predictions remain comparably
accurate.
|
Kernel Distillation for Fast Gaussian Processes Prediction | Gaussian processes (GPs) are flexible models that can capture complex
structure in large-scale dataset due to their non-parametric nature. However,
the usage of GPs in real-world application is limited due to their high
computational cost at inference time. In this paper, we introduce a new
framework, \textit{kernel distillation}, to approximate a fully trained teacher
GP model with kernel matrix of size $n\times n$ for $n$ training points. We
combine inducing points method with sparse low-rank approximation in the
distillation procedure. The distilled student GP model only costs $O(m^2)$
storage for $m$ inducing points where $m \ll n$ and improves the inference time
complexity. We demonstrate empirically that kernel distillation provides better
trade-off between the prediction time and the test performance compared to the
alternatives.
|
Nested LSTMs | We propose Nested LSTMs (NLSTM), a novel RNN architecture with multiple
levels of memory. Nested LSTMs add depth to LSTMs via nesting as opposed to
stacking. The value of a memory cell in an NLSTM is computed by an LSTM cell,
which has its own inner memory cell. Specifically, instead of computing the
value of the (outer) memory cell as $c^{outer}_t = f_t \odot c_{t-1} + i_t
\odot g_t$, NLSTM memory cells use the concatenation $(f_t \odot c_{t-1}, i_t
\odot g_t)$ as input to an inner LSTM (or NLSTM) memory cell, and set
$c^{outer}_t$ = $h^{inner}_t$. Nested LSTMs outperform both stacked and
single-layer LSTMs with similar numbers of parameters in our experiments on
various character-level language modeling tasks, and the inner memories of an
LSTM learn longer term dependencies compared with the higher-level units of a
stacked LSTM.
|
From BoW to CNN: Two Decades of Texture Representation for Texture
Classification | Texture is a fundamental characteristic of many types of images, and texture
representation is one of the essential and challenging problems in computer
vision and pattern recognition which has attracted extensive research
attention. Since 2000, texture representations based on Bag of Words (BoW) and
on Convolutional Neural Networks (CNNs) have been extensively studied with
impressive performance. Given this period of remarkable evolution, this paper
aims to present a comprehensive survey of advances in texture representation
over the last two decades. More than 200 major publications are cited in this
survey covering different aspects of the research, which includes (i) problem
description; (ii) recent advances in the broad categories of BoW-based,
CNN-based and attribute-based methods; and (iii) evaluation issues,
specifically benchmark datasets and state of the art results. In retrospect of
what has been achieved so far, the survey discusses open challenges and
directions for future research.
|
Deep Multi-view Learning to Rank | We study the problem of learning to rank from multiple information sources.
Though multi-view learning and learning to rank have been studied extensively
leading to a wide range of applications, multi-view learning to rank as a
synergy of both topics has received little attention. The aim of the paper is
to propose a composite ranking method while keeping a close correlation with
the individual rankings simultaneously. We present a generic framework for
multi-view subspace learning to rank (MvSL2R), and two novel solutions are
introduced under the framework. The first solution captures information of
feature mappings from within each view as well as across views using
autoencoder-like networks. Novel feature embedding methods are formulated in
the optimization of multi-view unsupervised and discriminant autoencoders.
Moreover, we introduce an end-to-end solution to learning towards both the
joint ranking objective and the individual rankings. The proposed solution
enhances the joint ranking with minimum view-specific ranking loss, so that it
can achieve the maximum global view agreements in a single optimization
process. The proposed method is evaluated on three different ranking problems,
i.e. university ranking, multi-view lingual text ranking and image data
ranking, providing superior results compared to related methods.
|
Pretraining Deep Actor-Critic Reinforcement Learning Algorithms With
Expert Demonstrations | Pretraining with expert demonstrations have been found useful in speeding up
the training process of deep reinforcement learning algorithms since less
online simulation data is required. Some people use supervised learning to
speed up the process of feature learning, others pretrain the policies by
imitating expert demonstrations. However, these methods are unstable and not
suitable for actor-critic reinforcement learning algorithms. Also, some
existing methods rely on the global optimum assumption, which is not true in
most scenarios. In this paper, we employ expert demonstrations in a
actor-critic reinforcement learning framework, and meanwhile ensure that the
performance is not affected by the fact that expert demonstrations are not
global optimal. We theoretically derive a method for computing policy gradients
and value estimators with only expert demonstrations. Our method is
theoretically plausible for actor-critic reinforcement learning algorithms that
pretrains both policy and value functions. We apply our method to two of the
typical actor-critic reinforcement learning algorithms, DDPG and ACER, and
demonstrate with experiments that our method not only outperforms the RL
algorithms without pretraining process, but also is more simulation efficient.
|
Learning from Informants: Relations between Learning Success Criteria | Learning from positive and negative information, so-called \emph{informants},
being one of the models for human and machine learning introduced by
E.~M.~Gold, is investigated. Particularly, naturally arising questions about
this learning setting, originating in results on learning from solely positive
information, are answered. By a carefully arranged argument learners can be
assumed to only change their hypothesis in case it is inconsistent with the
data (such a learning behavior is called \emph{conservative}). The deduced main
theorem states the relations between the most important delayable learning
success criteria, being the ones not ruined by a delayed in time hypothesis
output. Additionally, our investigations concerning the non-delayable
requirement of consistent learning underpin the claim for \emph{delayability}
being the right structural property to gain a deeper understanding concerning
the nature of learning success criteria. Moreover, we obtain an anomalous
\emph{hierarchy} when allowing for an increasing finite number of
\emph{anomalies} of the hypothesized language by the learner compared with the
language to be learned. In contrast to the vacillatory hierarchy for learning
from solely positive information, we observe a \emph{duality} depending on
whether infinitely many \emph{vacillations} between different (almost) correct
hypotheses are still considered a successful learning behavior.
|
Naive Bayes Entrapment Detection for Planetary Rovers | Entrapment detection is a prerequisite for planetary rovers to perform
autonomous rescue procedure. In this study, rover entrapment and approximated
entrapment criteria are formally defined. Entrapment detection using Naive
Bayes classifiers is proposed and discussed along with results from experiments
where the Naive Bayes entrapment detector is applied to AutoKralwer rovers. And
final conclusions and further discussions are presented in the final section.
|
Evaluating the Robustness of Neural Networks: An Extreme Value Theory
Approach | The robustness of neural networks to adversarial examples has received great
attention due to security implications. Despite various attack approaches to
crafting visually imperceptible adversarial examples, little has been developed
towards a comprehensive measure of robustness. In this paper, we provide a
theoretical justification for converting robustness analysis into a local
Lipschitz constant estimation problem, and propose to use the Extreme Value
Theory for efficient evaluation. Our analysis yields a novel robustness metric
called CLEVER, which is short for Cross Lipschitz Extreme Value for nEtwork
Robustness. The proposed CLEVER score is attack-agnostic and computationally
feasible for large neural networks. Experimental results on various networks,
including ResNet, Inception-v3 and MobileNet, show that (i) CLEVER is aligned
with the robustness indication measured by the $\ell_2$ and $\ell_\infty$ norms
of adversarial examples from powerful attacks, and (ii) defended networks using
defensive distillation or bounded ReLU indeed achieve better CLEVER scores. To
the best of our knowledge, CLEVER is the first attack-independent robustness
metric that can be applied to any neural network classifier.
|
DxNAT - Deep Neural Networks for Explaining Non-Recurring Traffic
Congestion | Non-recurring traffic congestion is caused by temporary disruptions, such as
accidents, sports games, adverse weather, etc. We use data related to real-time
traffic speed, jam factors (a traffic congestion indicator), and events
collected over a year from Nashville, TN to train a multi-layered deep neural
network. The traffic dataset contains over 900 million data records. The
network is thereafter used to classify the real-time data and identify
anomalous operations. Compared with traditional approaches of using statistical
or machine learning techniques, our model reaches an accuracy of 98.73 percent
when identifying traffic congestion caused by football games. Our approach
first encodes the traffic across a region as a scaled image. After that the
image data from different timestamps is fused with event- and time-related
data. Then a crossover operator is used as a data augmentation method to
generate training datasets with more balanced classes. Finally, we use the
receiver operating characteristic (ROC) analysis to tune the sensitivity of the
classifier. We present the analysis of the training time and the inference time
separately.
|
Deep Learning of Nonnegativity-Constrained Autoencoders for Enhanced
Understanding of Data | Unsupervised feature extractors are known to perform an efficient and
discriminative representation of data. Insight into the mappings they perform
and human ability to understand them, however, remain very limited. This is
especially prominent when multilayer deep learning architectures are used. This
paper demonstrates how to remove these bottlenecks within the architecture of
Nonnegativity Constrained Autoencoder (NCSAE). It is shown that by using both
L1 and L2 regularization that induce nonnegativity of weights, most of the
weights in the network become constrained to be nonnegative thereby resulting
into a more understandable structure with minute deterioration in
classification accuracy. Also, this proposed approach extracts features that
are more sparse and produces additional output layer sparsification. The method
is analyzed for accuracy and feature interpretation on the MNIST data, the NORB
normalized uniform object data, and the Reuters text categorization dataset.
|
A New Backpropagation Algorithm without Gradient Descent | The backpropagation algorithm, which had been originally introduced in the
1970s, is the workhorse of learning in neural networks. This backpropagation
algorithm makes use of the famous machine learning algorithm known as Gradient
Descent, which is a first-order iterative optimization algorithm for finding
the minimum of a function. To find a local minimum of a function using gradient
descent, one takes steps proportional to the negative of the gradient (or of
the approximate gradient) of the function at the current point. In this paper,
we develop an alternative to the backpropagation without the use of the
Gradient Descent Algorithm, but instead we are going to devise a new algorithm
to find the error in the weights and biases of an artificial neuron using
Moore-Penrose Pseudo Inverse. The numerical studies and the experiments
performed on various datasets are used to verify the working of this
alternative algorithm.
|
Fusarium Damaged Kernels Detection Using Transfer Learning on Deep
Neural Network Architecture | The present work shows the application of transfer learning for a pre-trained
deep neural network (DNN), using a small image dataset ($\approx$ 12,000) on a
single workstation with enabled NVIDIA GPU card that takes up to 1 hour to
complete the training task and archive an overall average accuracy of $94.7\%$.
The DNN presents a $20\%$ score of misclassification for an external test
dataset. The accuracy of the proposed methodology is equivalent to ones using
HSI methodology $(81\%-91\%)$ used for the same task, but with the advantage of
being independent on special equipment to classify wheat kernel for FHB
symptoms.
|
Incremental kernel PCA and the Nystr\"om method | Incremental versions of batch algorithms are often desired, for increased
time efficiency in the streaming data setting, or increased memory efficiency
in general. In this paper we present a novel algorithm for incremental kernel
PCA, based on rank one updates to the eigendecomposition of the kernel matrix,
which is more computationally efficient than comparable existing algorithms. We
extend our algorithm to incremental calculation of the Nystr\"om approximation
to the kernel matrix, the first such algorithm proposed. Incremental
calculation of the Nystr\"om approximation leads to further gains in memory
efficiency, and allows for empirical evaluation of when a subset of sufficient
size has been obtained.
|
Matrix completion with deterministic pattern - a geometric perspective | We consider the matrix completion problem with a deterministic pattern of
observed entries. In this setting, we aim to answer the question: under what
condition there will be (at least locally) unique solution to the matrix
completion problem, i.e., the underlying true matrix is identifiable. We answer
the question from a certain point of view and outline a geometric perspective.
We give an algebraically verifiable sufficient condition, which we call the
well-posedness condition, for the local uniqueness of MRMC solutions. We argue
that this condition is necessary for local stability of MRMC solutions, and we
show that the condition is generic using the characteristic rank. We also argue
that the low-rank approximation approaches are more stable than MRMC and
further propose a sequential statistical testing procedure to determine the
"true" rank from observed entries. Finally, we provide numerical examples aimed
at verifying validity of the presented theory.
|
Leveraging Adiabatic Quantum Computation for Election Forecasting | Accurate, reliable sampling from fully-connected graphs with arbitrary
correlations is a difficult problem. Such sampling requires knowledge of the
probabilities of observing every possible state of a graph. As graph size
grows, the number of model states becomes intractably large and efficient
computation requires full sampling be replaced with heuristics and algorithms
that are only approximations of full sampling. This work investigates the
potential impact of adiabatic quantum computation for sampling purposes,
building on recent successes training Boltzmann machines using a quantum
device. We investigate the use case of quantum computation to train Boltzmann
machines for predicting the 2016 Presidential election.
|
Optimizing Non-decomposable Measures with Deep Networks | We present a class of algorithms capable of directly training deep neural
networks with respect to large families of task-specific performance measures
such as the F-measure and the Kullback-Leibler divergence that are structured
and non-decomposable. This presents a departure from standard deep learning
techniques that typically use squared or cross-entropy loss functions (that are
decomposable) to train neural networks. We demonstrate that directly training
with task-specific loss functions yields much faster and more stable
convergence across problems and datasets. Our proposed algorithms and
implementations have several novel features including (i) convergence to first
order stationary points despite optimizing complex objective functions; (ii)
use of fewer training samples to achieve a desired level of convergence, (iii)
a substantial reduction in training time, and (iv) a seamless integration of
our implementation into existing symbolic gradient frameworks. We implement our
techniques on a variety of deep architectures including multi-layer perceptrons
and recurrent neural networks and show that on a variety of benchmark and real
data sets, our algorithms outperform traditional approaches to training deep
networks, as well as some recent approaches to task-specific training of neural
networks.
|
A Modified Sigma-Pi-Sigma Neural Network with Adaptive Choice of
Multinomials | Sigma-Pi-Sigma neural networks (SPSNNs) as a kind of high-order neural
networks can provide more powerful mapping capability than the traditional
feedforward neural networks (Sigma-Sigma neural networks). In the existing
literature, in order to reduce the number of the Pi nodes in the Pi layer, a
special multinomial P_s is used in SPSNNs. Each monomial in P_s is linear with
respect to each particular variable sigma_i when the other variables are taken
as constants. Therefore, the monomials like sigma_i^n or sigma_i^n sigma_j with
n>1 are not included. This choice may be somehow intuitive, but is not
necessarily the best. We propose in this paper a modified Sigma-Pi-Sigma neural
network (MSPSNN) with an adaptive approach to find a better multinomial for a
given problem. To elaborate, we start from a complete multinomial with a given
order. Then we employ a regularization technique in the learning process for
the given problem to reduce the number of monomials used in the multinomial,
and end up with a new SPSNN involving the same number of monomials (= the
number of nodes in the Pi-layer) as in P_s. Numerical experiments on some
benchmark problems show that our MSPSNN behaves better than the traditional
SPSNN with P_s.
|
Rethinking the Smaller-Norm-Less-Informative Assumption in Channel
Pruning of Convolution Layers | Model pruning has become a useful technique that improves the computational
efficiency of deep learning, making it possible to deploy solutions in
resource-limited scenarios. A widely-used practice in relevant work assumes
that a smaller-norm parameter or feature plays a less informative role at the
inference time. In this paper, we propose a channel pruning technique for
accelerating the computations of deep convolutional neural networks (CNNs) that
does not critically rely on this assumption. Instead, it focuses on direct
simplification of the channel-to-channel computation graph of a CNN without the
need of performing a computationally difficult and not-always-useful task of
making high-dimensional tensors of CNN structured sparse. Our approach takes
two stages: first to adopt an end-to- end stochastic training method that
eventually forces the outputs of some channels to be constant, and then to
prune those constant channels from the original neural network by adjusting the
biases of their impacting layers such that the resulting compact model can be
quickly fine-tuned. Our approach is mathematically appealing from an
optimization perspective and easy to reproduce. We experimented our approach
through several image learning benchmarks and demonstrate its interesting
aspects and competitive performance.
|
Distributed Newton Methods for Deep Neural Networks | Deep learning involves a difficult non-convex optimization problem with a
large number of weights between any two adjacent layers of a deep structure. To
handle large data sets or complicated networks, distributed training is needed,
but the calculation of function, gradient, and Hessian is expensive. In
particular, the communication and the synchronization cost may become a
bottleneck. In this paper, we focus on situations where the model is
distributedly stored, and propose a novel distributed Newton method for
training deep neural networks. By variable and feature-wise data partitions,
and some careful designs, we are able to explicitly use the Jacobian matrix for
matrix-vector products in the Newton method. Some techniques are incorporated
to reduce the running time as well as the memory consumption. First, to reduce
the communication cost, we propose a diagonalization method such that an
approximate Newton direction can be obtained without communication between
machines. Second, we consider subsampled Gauss-Newton matrices for reducing the
running time as well as the communication cost. Third, to reduce the
synchronization cost, we terminate the process of finding an approximate Newton
direction even though some nodes have not finished their tasks. Details of some
implementation issues in distributed environments are thoroughly investigated.
Experiments demonstrate that the proposed method is effective for the
distributed training of deep neural networks. In compared with stochastic
gradient methods, it is more robust and may give better test accuracy.
|
Alternating Multi-bit Quantization for Recurrent Neural Networks | Recurrent neural networks have achieved excellent performance in many
applications. However, on portable devices with limited resources, the models
are often too large to deploy. For applications on the server with large scale
concurrent requests, the latency during inference can also be very critical for
costly computing resources. In this work, we address these problems by
quantizing the network, both weights and activations, into multiple binary
codes {-1,+1}. We formulate the quantization as an optimization problem. Under
the key observation that once the quantization coefficients are fixed the
binary codes can be derived efficiently by binary search tree, alternating
minimization is then applied. We test the quantization for two well-known RNNs,
i.e., long short term memory (LSTM) and gated recurrent unit (GRU), on the
language models. Compared with the full-precision counter part, by 2-bit
quantization we can achieve ~16x memory saving and ~6x real inference
acceleration on CPUs, with only a reasonable loss in the accuracy. By 3-bit
quantization, we can achieve almost no loss in the accuracy or even surpass the
original model, with ~10.5x memory saving and ~3x real inference acceleration.
Both results beat the exiting quantization works with large margins. We extend
our alternating quantization to image classification tasks. In both RNNs and
feedforward neural networks, the method also achieves excellent performance.
|
Bootstrapping and Multiple Imputation Ensemble Approaches for Missing
Data | Presence of missing values in a dataset can adversely affect the performance
of a classifier. Single and Multiple Imputation are normally performed to fill
in the missing values. In this paper, we present several variants of combining
single and multiple imputation with bootstrapping to create ensembles that can
model uncertainty and diversity in the data, and that are robust to high
missingness in the data. We present three ensemble strategies: bootstrapping on
incomplete data followed by (i) single imputation and (ii) multiple imputation,
and (iii) multiple imputation ensemble without bootstrapping. We perform an
extensive evaluation of the performance of the these ensemble strategies on 8
datasets by varying the missingness ratio. Our results show that bootstrapping
followed by multiple imputation using expectation maximization is the most
robust method even at high missingness ratio (up to 30%). For small missingness
ratio (up to 10%) most of the ensemble methods perform quivalently but better
than single imputation. Kappa-error plots suggest that accurate classifiers
with reasonable diversity is the reason for this behaviour. A consistent
observation in all the datasets suggests that for small missingness (up to
10%), bootstrapping on incomplete data without any imputation produces
equivalent results to other ensemble methods.
|
Deep Neural Nets with Interpolating Function as Output Activation | We replace the output layer of deep neural nets, typically the softmax
function, by a novel interpolating function. And we propose end-to-end training
and testing algorithms for this new architecture. Compared to classical neural
nets with softmax function as output activation, the surrogate with
interpolating function as output activation combines advantages of both deep
and manifold learning. The new framework demonstrates the following major
advantages: First, it is better applicable to the case with insufficient
training data. Second, it significantly improves the generalization accuracy on
a wide variety of networks. The algorithm is implemented in PyTorch, and code
will be made publicly available.
|
Augmented Space Linear Model | The linear model uses the space defined by the input to project the target or
desired signal and find the optimal set of model parameters. When the problem
is nonlinear, the adaption requires nonlinear models for good performance, but
it becomes slower and more cumbersome. In this paper, we propose a linear model
called Augmented Space Linear Model (ASLM), which uses the full joint space of
input and desired signal as the projection space and approaches the performance
of nonlinear models. This new algorithm takes advantage of the linear solution,
and corrects the estimate for the current testing phase input with the error
assigned to the input space neighborhood in the training phase. This algorithm
can solve the nonlinear problem with the computational efficiency of linear
methods, which can be regarded as a trade off between accuracy and
computational complexity. Making full use of the training data, the proposed
augmented space model may provide a new way to improve many modeling tasks.
|
Clustering and Unsupervised Anomaly Detection with L2 Normalized Deep
Auto-Encoder Representations | Clustering is essential to many tasks in pattern recognition and computer
vision. With the advent of deep learning, there is an increasing interest in
learning deep unsupervised representations for clustering analysis. Many works
on this domain rely on variants of auto-encoders and use the encoder outputs as
representations/features for clustering. In this paper, we show that an l2
normalization constraint on these representations during auto-encoder training,
makes the representations more separable and compact in the Euclidean space
after training. This greatly improves the clustering accuracy when k-means
clustering is employed on the representations. We also propose a clustering
based unsupervised anomaly detection method using l2 normalized deep
auto-encoder representations. We show the effect of l2 normalization on anomaly
detection accuracy. We further show that the proposed anomaly detection method
greatly improves accuracy compared to previously proposed deep methods such as
reconstruction error based anomaly detection.
|
Training Neural Networks by Using Power Linear Units (PoLUs) | In this paper, we introduce "Power Linear Unit" (PoLU) which increases the
nonlinearity capacity of a neural network and thus helps improving its
performance. PoLU adopts several advantages of previously proposed activation
functions. First, the output of PoLU for positive inputs is designed to be
identity to avoid the gradient vanishing problem. Second, PoLU has a non-zero
output for negative inputs such that the output mean of the units is close to
zero, hence reducing the bias shift effect. Thirdly, there is a saturation on
the negative part of PoLU, which makes it more noise-robust for negative
inputs. Furthermore, we prove that PoLU is able to map more portions of every
layer's input to the same space by using the power function and thus increases
the number of response regions of the neural network. We use image
classification for comparing our proposed activation function with others. In
the experiments, MNIST, CIFAR-10, CIFAR-100, Street View House Numbers (SVHN)
and ImageNet are used as benchmark datasets. The neural networks we implemented
include widely-used ELU-Network, ResNet-50, and VGG16, plus a couple of shallow
networks. Experimental results show that our proposed activation function
outperforms other state-of-the-art models with most networks.
|
On Polynomial time Constructions of Minimum Height Decision Tree | In this paper we study a polynomial time algorithms that for an input
$A\subseteq {B_m}$ outputs a decision tree for $A$ of minimum depth. This
problem has many applications that include, to name a few, computer vision,
group testing, exact learning from membership queries and game theory.
Arkin et al. and Moshkov gave a polynomial time $(\ln |A|)$- approximation
algorithm (for the depth). The result of Dinur and Steurer for set cover
implies that this problem cannot be approximated with ratio $(1-o(1))\cdot \ln
|A|$, unless P=NP. Moskov the combinatorial measure of extended teaching
dimension of $A$, $ETD(A)$. He showed that $ETD(A)$ is a lower bound for the
depth of the decision tree for $A$ and then gave an {\it exponential time}
$ETD(A)/\log(ETD(A))$-approximation algorithm.
In this paper we further study the $ETD(A)$ measure and a new combinatorial
measure, $DEN(A)$, that we call the density of the set $A$. We show that
$DEN(A)\le ETD(A)+1$. We then give two results. The first result is that the
lower bound $ETD(A)$ of Moshkov for the depth of the decision tree for $A$ is
greater than the bounds that are obtained by the classical technique used in
the literature. The second result is a polynomial time $(\ln 2)
DEN(A)$-approximation (and therefore $(\ln 2) ETD(A)$-approximation) algorithm
for the depth of the decision tree of $A$. We also show that a better
approximation ratio implies P=NP.
We then apply the above results to learning the class of disjunctions of
predicates from membership queries. We show that the $ETD$ of this class is
bounded from above by the degree $d$ of its Hasse diagram. We then show that
Moshkov algorithm can be run in polynomial time and is $(d/\log
d)$-approximation algorithm. This gives optimal algorithms when the degree is
constant. For example, learning axis parallel rays over constant dimension
space.
|
A Nonparametric Delayed Feedback Model for Conversion Rate Prediction | Predicting conversion rates (CVRs) in display advertising (e.g., predicting
the proportion of users who purchase an item (i.e., a conversion) after its
corresponding ad is clicked) is important when measuring the effects of ads
shown to users and to understanding the interests of the users. There is
generally a time delay (i.e., so-called {\it delayed feedback}) between the ad
click and conversion. Owing to the delayed feedback, samples that are converted
after an observation period may be treated as negative. To overcome this
drawback, CVR prediction assuming that the time delay follows an exponential
distribution has been proposed. In practice, however, there is no guarantee
that the delay is generated from the exponential distribution, and the best
distribution with which to represent the delay depends on the data. In this
paper, we propose a nonparametric delayed feedback model for CVR prediction
that represents the distribution of the time delay without assuming a
parametric distribution, such as an exponential or Weibull distribution.
Because the distribution of the time delay is modeled depending on the content
of an ad and the features of a user, various shapes of the distribution can be
represented potentially. In experiments, we show that the proposed model can
capture the distribution for the time delay on a synthetic dataset, even when
the distribution is complicated. Moreover, on a real dataset, we show that the
proposed model outperforms the existing method that assumes an exponential
distribution for the time delay in terms of conversion rate prediction.
|
VR-Goggles for Robots: Real-to-sim Domain Adaptation for Visual Control | In this paper, we deal with the reality gap from a novel perspective,
targeting transferring Deep Reinforcement Learning (DRL) policies learned in
simulated environments to the real-world domain for visual control tasks.
Instead of adopting the common solutions to the problem by increasing the
visual fidelity of synthetic images output from simulators during the training
phase, we seek to tackle the problem by translating the real-world image
streams back to the synthetic domain during the deployment phase, to make the
robot feel at home. We propose this as a lightweight, flexible, and efficient
solution for visual control, as 1) no extra transfer steps are required during
the expensive training of DRL agents in simulation; 2) the trained DRL agents
will not be constrained to being deployable in only one specific real-world
environment; 3) the policy training and the transfer operations are decoupled,
and can be conducted in parallel. Besides this, we propose a simple yet
effective shift loss that is agnostic to the downstream task, to constrain the
consistency between subsequent frames which is important for consistent policy
outputs. We validate the shift loss for artistic style transfer for videos and
domain adaptation, and validate our visual control approach in indoor and
outdoor robotics experiments.
|
ChronoNet: A Deep Recurrent Neural Network for Abnormal EEG
Identification | Brain-related disorders such as epilepsy can be diagnosed by analyzing
electroencephalograms (EEG). However, manual analysis of EEG data requires
highly trained clinicians, and is a procedure that is known to have relatively
low inter-rater agreement (IRA). Moreover, the volume of the data and the rate
at which new data becomes available make manual interpretation a
time-consuming, resource-hungry, and expensive process. In contrast, automated
analysis of EEG data offers the potential to improve the quality of patient
care by shortening the time to diagnosis and reducing manual error. In this
paper, we focus on one of the first steps in interpreting an EEG session -
identifying whether the brain activity is abnormal or normal. To solve this
task, we propose a novel recurrent neural network (RNN) architecture termed
ChronoNet which is inspired by recent developments from the field of image
classification and designed to work efficiently with EEG data. ChronoNet is
formed by stacking multiple 1D convolution layers followed by deep gated
recurrent unit (GRU) layers where each 1D convolution layer uses multiple
filters of exponentially varying lengths and the stacked GRU layers are densely
connected in a feed-forward manner. We used the recently released TUH Abnormal
EEG Corpus dataset for evaluating the performance of ChronoNet. Unlike previous
studies using this dataset, ChronoNet directly takes time-series EEG as input
and learns meaningful representations of brain activity patterns. ChronoNet
outperforms the previously reported best results by 7.79% thereby setting a new
benchmark for this dataset. Furthermore, we demonstrate the domain-independent
nature of ChronoNet by successfully applying it to classify speech commands.
|
One-class Collective Anomaly Detection based on Long Short-Term Memory
Recurrent Neural Networks | Intrusion detection for computer network systems has been becoming one of the
most critical tasks for network administrators today. It has an important role
for organizations, governments and our society due to the valuable resources
hosted on computer networks. Traditional misuse detection strategies are unable
to detect new and unknown intrusion types. In contrast, anomaly detection in
network security aims to distinguish between illegal or malicious events and
normal behavior of network systems. Anomaly detection can be considered as a
classification problem where it builds models of normal network behavior, of
which it uses to detect new patterns that significantly deviate from the model.
Most of the current approaches on anomaly detection is based on the learning of
normal behavior and anomalous actions. They do not include memory that is they
do not take into account previous events classify new ones. In this paper, we
propose a one class collective anomaly detection model based on neural network
learning. Normally a Long Short Term Memory Recurrent Neural Network (LSTM RNN)
is trained only on normal data, and it is capable of predicting several time
steps ahead of an input. In our approach, a LSTM RNN is trained on normal time
series data before performing a prediction for each time step. Instead of
considering each time-step separately, the observation of prediction errors
from a certain number of time-steps is now proposed as a new idea for detecting
collective anomalies. The prediction errors of a certain number of the latest
time-steps above a threshold will indicate a collective anomaly. The model is
evaluated on a time series version of the KDD 1999 dataset. The experiments
demonstrate that the proposed model is capable to detect collective anomaly
efficiently
|
Elements of Effective Deep Reinforcement Learning towards Tactical
Driving Decision Making | Tactical driving decision making is crucial for autonomous driving systems
and has attracted considerable interest in recent years. In this paper, we
propose several practical components that can speed up deep reinforcement
learning algorithms towards tactical decision making tasks: 1) non-uniform
action skipping as a more stable alternative to action-repetition frame
skipping, 2) a counter-based penalty for lanes on which ego vehicle has less
right-of-road, and 3) heuristic inference-time action masking for apparently
undesirable actions. We evaluate the proposed components in a realistic driving
simulator and compare them with several baselines. Results show that the
proposed scheme provides superior performance in terms of safety, efficiency,
and comfort.
|
Classifying medical notes into standard disease codes using Machine
Learning | We investigate the automatic classification of patient discharge notes into
standard disease labels. We find that Convolutional Neural Networks with
Attention outperform previous algorithms used in this task, and suggest further
areas for improvement.
|
Dense 3D Object Reconstruction from a Single Depth View | In this paper, we propose a novel approach, 3D-RecGAN++, which reconstructs
the complete 3D structure of a given object from a single arbitrary depth view
using generative adversarial networks. Unlike existing work which typically
requires multiple views of the same object or class labels to recover the full
3D geometry, the proposed 3D-RecGAN++ only takes the voxel grid representation
of a depth view of the object as input, and is able to generate the complete 3D
occupancy grid with a high resolution of 256^3 by recovering the
occluded/missing regions. The key idea is to combine the generative
capabilities of autoencoders and the conditional Generative Adversarial
Networks (GAN) framework, to infer accurate and fine-grained 3D structures of
objects in high-dimensional voxel space. Extensive experiments on large
synthetic datasets and real-world Kinect datasets show that the proposed
3D-RecGAN++ significantly outperforms the state of the art in single view 3D
object reconstruction, and is able to reconstruct unseen types of objects.
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.