title
stringlengths 7
246
| abstract
stringlengths 6
3.31k
|
|---|---|
QSAR Classification Modeling for Bioactivity of Molecular Structure via
SPL-Logsum | Quantitative structure-activity relationship (QSAR) modelling is effective
'bridge' to search the reliable relationship related bioactivity to molecular
structure. A QSAR classification model contains a lager number of redundant,
noisy and irrelevant descriptors. To address this problem, various of methods
have been proposed for descriptor selection. Generally, they can be grouped
into three categories: filters, wrappers, and embedded methods. Regularization
method is an important embedded technology, which can be used for continuous
shrinkage and automatic descriptors selection. In recent years, the interest of
researchers in the application of regularization techniques is increasing in
descriptors selection , such as, logistic regression(LR) with $L_1$ penalty. In
this paper, we proposed a novel descriptor selection method based on self-paced
learning(SPL) with Logsum penalized LR for predicting the bioactivity of
molecular structure. SPL inspired by the learning process of humans and animals
that gradually learns from easy samples(smaller losses) to hard samples(bigger
losses) samples into training and Logsum regularization has capacity to select
few meaningful and significant molecular descriptors, respectively.
Experimental results on simulation and three public QSAR datasets show that our
proposed SPL-Logsum method outperforms other commonly used sparse methods in
terms of classification performance and model interpretation.
|
Distributed Distributional Deterministic Policy Gradients | This work adopts the very successful distributional perspective on
reinforcement learning and adapts it to the continuous control setting. We
combine this within a distributed framework for off-policy learning in order to
develop what we call the Distributed Distributional Deep Deterministic Policy
Gradient algorithm, D4PG. We also combine this technique with a number of
additional, simple improvements such as the use of $N$-step returns and
prioritized experience replay. Experimentally we examine the contribution of
each of these individual components, and show how they interact, as well as
their combined contributions. Our results show that across a wide variety of
simple control tasks, difficult manipulation tasks, and a set of hard
obstacle-based locomotion tasks the D4PG algorithm achieves state of the art
performance.
|
State Distribution-aware Sampling for Deep Q-learning | A critical and challenging problem in reinforcement learning is how to learn
the state-action value function from the experience replay buffer and
simultaneously keep sample efficiency and faster convergence to a high quality
solution. In prior works, transitions are uniformly sampled at random from the
replay buffer or sampled based on their priority measured by
temporal-difference (TD) error. However, these approaches do not fully take
into consideration the intrinsic characteristics of transition distribution in
the state space and could result in redundant and unnecessary TD updates,
slowing down the convergence of the learning procedure. To overcome this
problem, we propose a novel state distribution-aware sampling method to balance
the replay times for transitions with skew distribution, which takes into
account both the occurrence frequencies of transitions and the uncertainty of
state-action values. Consequently, our approach could reduce the unnecessary TD
updates and increase the TD updates for state-action value with more
uncertainty, making the experience replay more effective and efficient.
Extensive experiments are conducted on both classic control tasks and Atari
2600 games based on OpenAI gym platform and the experimental results
demonstrate the effectiveness of our approach in comparison with the standard
DQN approach.
|
Quantum generative adversarial networks | Quantum machine learning is expected to be one of the first potential
general-purpose applications of near-term quantum devices. A major recent
breakthrough in classical machine learning is the notion of generative
adversarial training, where the gradients of a discriminator model are used to
train a separate generative model. In this work and a companion paper, we
extend adversarial training to the quantum domain and show how to construct
generative adversarial networks using quantum circuits. Furthermore, we also
show how to compute gradients -- a key element in generative adversarial
network training -- using another quantum circuit. We give an example of a
simple practical circuit ansatz to parametrize quantum machine learning models
and perform a simple numerical experiment to demonstrate that quantum
generative adversarial networks can be trained successfully.
|
A Theory of Statistical Inference for Ensuring the Robustness of
Scientific Results | Inference is the process of using facts we know to learn about facts we do
not know. A theory of inference gives assumptions necessary to get from the
former to the latter, along with a definition for and summary of the resulting
uncertainty. Any one theory of inference is neither right nor wrong, but merely
an axiom that may or may not be useful. Each of the many diverse theories of
inference can be valuable for certain applications. However, no existing theory
of inference addresses the tendency to choose, from the range of plausible data
analysis specifications consistent with prior evidence, those that
inadvertently favor one's own hypotheses. Since the biases from these choices
are a growing concern across scientific fields, and in a sense the reason the
scientific community was invented in the first place, we introduce a new theory
of inference designed to address this critical problem. We introduce hacking
intervals, which are the range of a summary statistic one may obtain given a
class of possible endogenous manipulations of the data. Hacking intervals
require no appeal to hypothetical data sets drawn from imaginary
superpopulations. A scientific result with a small hacking interval is more
robust to researcher manipulation than one with a larger interval, and is often
easier to interpret than a classical confidence interval. Some versions of
hacking intervals turn out to be equivalent to classical confidence intervals,
which means they may also provide a more intuitive and potentially more useful
interpretation of classical confidence intervals.
|
Boltzmann Encoded Adversarial Machines | Restricted Boltzmann Machines (RBMs) are a class of generative neural network
that are typically trained to maximize a log-likelihood objective function. We
argue that likelihood-based training strategies may fail because the objective
does not sufficiently penalize models that place a high probability in regions
where the training data distribution has low probability. To overcome this
problem, we introduce Boltzmann Encoded Adversarial Machines (BEAMs). A BEAM is
an RBM trained against an adversary that uses the hidden layer activations of
the RBM to discriminate between the training data and the probability
distribution generated by the model. We present experiments demonstrating that
BEAMs outperform RBMs and GANs on multiple benchmarks.
|
Crawling in Rogue's dungeons with (partitioned) A3C | Rogue is a famous dungeon-crawling video-game of the 80ies, the ancestor of
its gender. Rogue-like games are known for the necessity to explore partially
observable and always different randomly-generated labyrinths, preventing any
form of level replay. As such, they serve as a very natural and challenging
task for reinforcement learning, requiring the acquisition of complex,
non-reactive behaviors involving memory and planning. In this article we show
how, exploiting a version of A3C partitioned on different situations, the agent
is able to reach the stairs and descend to the next level in 98% of cases.
|
All-Optical Machine Learning Using Diffractive Deep Neural Networks | We introduce an all-optical Diffractive Deep Neural Network (D2NN)
architecture that can learn to implement various functions after deep
learning-based design of passive diffractive layers that work collectively. We
experimentally demonstrated the success of this framework by creating
3D-printed D2NNs that learned to implement handwritten digit classification and
the function of an imaging lens at terahertz spectrum. With the existing
plethora of 3D-printing and other lithographic fabrication methods as well as
spatial-light-modulators, this all-optical deep learning framework can perform,
at the speed of light, various complex functions that computer-based neural
networks can implement, and will find applications in all-optical image
analysis, feature detection and object classification, also enabling new camera
designs and optical components that can learn to perform unique tasks using
D2NNs.
|
A machine learning model for identifying cyclic alternating patterns in
the sleeping brain | Electroencephalography (EEG) is a method to record the electrical signals in
the brain. Recognizing the EEG patterns in the sleeping brain gives insights
into the understanding of sleeping disorders. The dataset under consideration
contains EEG data points associated with various physiological conditions. This
study attempts to generalize the detection of particular patterns associated
with the Non-Rapid Eye Movement (NREM) sleep cycle of the brain using a machine
learning model. The proposed model uses additional feature engineering to
incorporate sequential information for training a classifier to predict the
occurrence of Cyclic Alternating Pattern (CAP) sequences in the sleep cycle,
which are often associated with sleep disorders.
|
Neural-Brane: Neural Bayesian Personalized Ranking for Attributed
Network Embedding | Network embedding methodologies, which learn a distributed vector
representation for each vertex in a network, have attracted considerable
interest in recent years. Existing works have demonstrated that vertex
representation learned through an embedding method provides superior
performance in many real-world applications, such as node classification, link
prediction, and community detection. However, most of the existing methods for
network embedding only utilize topological information of a vertex, ignoring a
rich set of nodal attributes (such as, user profiles of an online social
network, or textual contents of a citation network), which is abundant in all
real-life networks. A joint network embedding that takes into account both
attributional and relational information entails a complete network information
and could further enrich the learned vector representations. In this work, we
present Neural-Brane, a novel Neural Bayesian Personalized Ranking based
Attributed Network Embedding. For a given network, Neural-Brane extracts latent
feature representation of its vertices using a designed neural network model
that unifies network topological information and nodal attributes; Besides, it
utilizes Bayesian personalized ranking objective, which exploits the proximity
ordering between a similar node-pair and a dissimilar node-pair. We evaluate
the quality of vertex embedding produced by Neural-Brane by solving the node
classification and clustering tasks on four real-world datasets. Experimental
results demonstrate the superiority of our proposed method over the
state-of-the-art existing methods.
|
Query-Efficient Black-Box Attack Against Sequence-Based Malware
Classifiers | In this paper, we present a generic, query-efficient black-box attack against
API call-based machine learning malware classifiers. We generate adversarial
examples by modifying the malware's API call sequences and non-sequential
features (printable strings), and these adversarial examples will be
misclassified by the target malware classifier without affecting the malware's
functionality. In contrast to previous studies, our attack minimizes the number
of malware classifier queries required. In addition, in our attack, the
attacker must only know the class predicted by the malware classifier; attacker
knowledge of the malware classifier's confidence score is optional. We evaluate
the attack effectiveness when attacks are performed against a variety of
malware classifier architectures, including recurrent neural network (RNN)
variants, deep neural networks, support vector machines, and gradient boosted
decision trees. Our attack success rate is around 98% when the classifier's
confidence score is known and 64% when just the classifier's predicted class is
known. We implement four state-of-the-art query-efficient attacks and show that
our attack requires fewer queries and less knowledge about the attacked model's
architecture than other existing query-efficient attacks, making it practical
for attacking cloud-based malware classifiers at a minimal cost.
|
Towards Dependable Deep Convolutional Neural Networks (CNNs) with
Out-distribution Learning | Detection and rejection of adversarial examples in security sensitive and
safety-critical systems using deep CNNs is essential. In this paper, we propose
an approach to augment CNNs with out-distribution learning in order to reduce
misclassification rate by rejecting adversarial examples. We empirically show
that our augmented CNNs can either reject or classify correctly most
adversarial examples generated using well-known methods ( >95% for MNIST and
>75% for CIFAR-10 on average). Furthermore, we achieve this without requiring
to train using any specific type of adversarial examples and without
sacrificing the accuracy of models on clean samples significantly (< 4%).
|
Block-Structure Based Time-Series Models For Graph Sequences | Although the computational and statistical trade-off for modeling single
graphs, for instance, using block models is relatively well understood,
extending such results to sequences of graphs has proven to be difficult. In
this work, we take a step in this direction by proposing two models for graph
sequences that capture: (a) link persistence between nodes across time, and (b)
community persistence of each node across time. In the first model, we assume
that the latent community of each node does not change over time, and in the
second model we relax this assumption suitably. For both of these proposed
models, we provide statistically and computationally efficient inference
algorithms, whose unique feature is that they leverage community detection
methods that work on single graphs. We also provide experimental results
validating the suitability of our models and methods on synthetic and real
instances.
|
Structured SUMCOR Multiview Canonical Correlation Analysis for
Large-Scale Data | The sum-of-correlations (SUMCOR) formulation of generalized canonical
correlation analysis (GCCA) seeks highly correlated low-dimensional
representations of different views via maximizing pairwise latent similarity of
the views. SUMCOR is considered arguably the most natural extension of
classical two-view CCA to the multiview case, and thus has numerous
applications in signal processing and data analytics. Recent work has proposed
effective algorithms for handling the SUMCOR problem at very large scale.
However, the existing scalable algorithms cannot incorporate structural
regularization and prior information -- which are critical for good performance
in real-world applications. In this work, we propose a new computational
framework for large-scale SUMCOR GCCA that can easily incorporate a suite of
structural regularizers which are frequently used in data analytics. The
updates of the proposed algorithm are lightweight and the memory complexity is
also low. In addition, the proposed algorithm can be readily implemented in a
parallel fashion. We show that the proposed algorithm converges to a
Karush-Kuhn-Tucker (KKT) point of the regularized SUMCOR problem. Judiciously
designed simulations and real-data experiments are employed to demonstrate the
effectiveness of the proposed algorithm.
|
Learning Manifolds from Non-stationary Streaming Data | Streaming adaptations of manifold learning based dimensionality reduction
methods, such as Isomap, are based on the assumption that a small initial batch
of observations is enough for exact learning of the manifold, while remaining
streaming data instances can be cheaply mapped to this manifold. However, there
are no theoretical results to show that this core assumption is valid.
Moreover, such methods typically assume that the underlying data distribution
is stationary. Such methods are not equipped to detect, or handle, sudden
changes or gradual drifts in the distribution that may occur when the data is
streaming. We present theoretical results to show that the quality of a
manifold asymptotically converges as the size of data increases. We then show
that a Gaussian Process Regression (GPR) model, that uses a manifold-specific
kernel function and is trained on an initial batch of sufficient size, can
closely approximate the state-of-art streaming Isomap algorithms. The
predictive variance obtained from the GPR prediction is then shown to be an
effective detector of changes in the underlying data distribution. Results on
several synthetic and real data sets show that the resulting algorithm can
effectively learn lower dimensional representation of high dimensional data in
a streaming setting, while identifying shifts in the generative distribution.
|
Measuring the Intrinsic Dimension of Objective Landscapes | Many recently trained neural networks employ large numbers of parameters to
achieve good performance. One may intuitively use the number of parameters
required as a rough gauge of the difficulty of a problem. But how accurate are
such notions? How many parameters are really needed? In this paper we attempt
to answer this question by training networks not in their native parameter
space, but instead in a smaller, randomly oriented subspace. We slowly increase
the dimension of this subspace, note at which dimension solutions first appear,
and define this to be the intrinsic dimension of the objective landscape. The
approach is simple to implement, computationally tractable, and produces
several suggestive conclusions. Many problems have smaller intrinsic dimensions
than one might suspect, and the intrinsic dimension for a given dataset varies
little across a family of models with vastly different sizes. This latter
result has the profound implication that once a parameter space is large enough
to solve a problem, extra parameters serve directly to increase the
dimensionality of the solution manifold. Intrinsic dimension allows some
quantitative comparison of problem difficulty across supervised, reinforcement,
and other types of learning where we conclude, for example, that solving the
inverted pendulum problem is 100 times easier than classifying digits from
MNIST, and playing Atari Pong from pixels is about as hard as classifying
CIFAR-10. In addition to providing new cartography of the objective landscapes
wandered by parameterized models, the method is a simple technique for
constructively obtaining an upper bound on the minimum description length of a
solution. A byproduct of this construction is a simple approach for compressing
networks, in some cases by more than 100 times.
|
Learning Software Constraints via Installation Attempts | Modern software systems are expected to be secure and contain all the latest
features, even when new versions of software are released multiple times an
hour. Each system may include many interacting packages. The problem of
installing multiple dependent packages has been extensively studied in the
past, yielding some promising solutions that work well in practice. However,
these assume that the developers declare all the dependencies and conflicts
between the packages. Oftentimes, the entire repository structure may not be
known upfront, for example when packages are developed by different vendors. In
this paper, we present algorithms for learning dependencies, conflicts and
defective packages from installation attempts. Our algorithms use combinatorial
data structures to generate queries that test installations and discover the
entire dependency structure. A query that the algorithms make corresponds to
trying to install a subset of packages and getting a Boolean feedback on
whether all constraints were satisfied in this subset. Our goal is to minimize
the query complexity of the algorithms. We prove lower and upper bounds on the
number of queries that these algorithms require to make for different settings
of the problem.
|
Deep Neural Network Based Subspace Learning of Robotic Manipulator
Workspace Mapping | The manipulator workspace mapping is an important problem in robotics and has
attracted significant attention in the community. However, most of the
pre-existing algorithms have expensive time complexity due to the reliance on
sophisticated kinematic equations. To solve this problem, this paper introduces
subspace learning (SL), a variant of subspace embedding, where a set of robot
and scope parameters is mapped to the corresponding workspace by a deep neural
network (DNN). Trained on a large dataset of around $\mathbf{6\times 10^4}$
samples obtained from a MATLAB$^\circledR$ implementation of a classical method
and sampling of designed uniform distributions, the experiments demonstrate
that the embedding significantly reduces run-time from $\mathbf{5.23 \times
10^3}$ s of traditional discretization method to $\mathbf{0.224}$ s, with high
accuracies (average F-measure is $\mathbf{0.9665}$ with batch gradient descent
and resilient backpropagation).
|
Estimate and Replace: A Novel Approach to Integrating Deep Neural
Networks with Existing Applications | Existing applications include a huge amount of knowledge that is out of reach
for deep neural networks. This paper presents a novel approach for integrating
calls to existing applications into deep learning architectures. Using this
approach, we estimate each application's functionality with an estimator, which
is implemented as a deep neural network (DNN). The estimator is then embedded
into a base network that we direct into complying with the application's
interface during an end-to-end optimization process. At inference time, we
replace each estimator with its existing application counterpart and let the
base network solve the task by interacting with the existing application. Using
this 'Estimate and Replace' method, we were able to train a DNN end-to-end with
less data and outperformed a matching DNN that did not interact with the
external application.
|
Developing a machine learning framework for estimating soil moisture
with VNIR hyperspectral data | In this paper, we investigate the potential of estimating the soil-moisture
content based on VNIR hyperspectral data combined with LWIR data. Measurements
from a multi-sensor field campaign represent the benchmark dataset which
contains measured hyperspectral, LWIR, and soil-moisture data conducted on
grassland site. We introduce a regression framework with three steps consisting
of feature selection, preprocessing, and well-chosen regression models. The
latter are mainly supervised machine learning models. An exception are the
self-organizing maps which combine unsupervised and supervised learning. We
analyze the impact of the distinct preprocessing methods on the regression
results. Of all regression models, the extremely randomized trees model without
preprocessing provides the best estimation performance. Our results reveal the
potential of the respective regression framework combined with the VNIR
hyperspectral data to estimate soil moisture measured under real-world
conditions. In conclusion, the results of this paper provide a basis for
further improvements in different research directions.
|
An Information-Theoretic View for Deep Learning | Deep learning has transformed computer vision, natural language processing,
and speech recognition\cite{badrinarayanan2017segnet, dong2016image,
ren2017faster, ji20133d}. However, two critical questions remain obscure: (1)
why do deep neural networks generalize better than shallow networks; and (2)
does it always hold that a deeper network leads to better performance?
Specifically, letting $L$ be the number of convolutional and pooling layers in
a deep neural network, and $n$ be the size of the training sample, we derive an
upper bound on the expected generalization error for this network, i.e.,
\begin{eqnarray*}
\mathbb{E}[R(W)-R_S(W)] \leq
\exp{\left(-\frac{L}{2}\log{\frac{1}{\eta}}\right)}\sqrt{\frac{2\sigma^2}{n}I(S,W)
}
\end{eqnarray*} where $\sigma >0$ is a constant depending on the loss
function, $0<\eta<1$ is a constant depending on the information loss for each
convolutional or pooling layer, and $I(S, W)$ is the mutual information between
the training sample $S$ and the output hypothesis $W$. This upper bound shows
that as the number of convolutional and pooling layers $L$ increases in the
network, the expected generalization error will decrease exponentially to zero.
Layers with strict information loss, such as the convolutional layers, reduce
the generalization error for the whole network; this answers the first
question. However, algorithms with zero expected generalization error does not
imply a small test error or $\mathbb{E}[R(W)]$. This is because
$\mathbb{E}[R_S(W)]$ is large when the information for fitting the data is lost
as the number of layers increases. This suggests that the claim `the deeper the
better' is conditioned on a small training error or $\mathbb{E}[R_S(W)]$.
Finally, we show that deep learning satisfies a weak notion of stability and
the sample complexity of deep neural networks will decrease as $L$ increases.
|
Efficient Multi-objective Neural Architecture Search via Lamarckian
Evolution | Neural Architecture Search aims at automatically finding neural architectures
that are competitive with architectures designed by human experts. While recent
approaches have achieved state-of-the-art predictive performance for image
recognition, they are problematic under resource constraints for two reasons:
(1)the neural architectures found are solely optimized for high predictive
performance, without penalizing excessive resource consumption, (2) most
architecture search methods require vast computational resources. We address
the first shortcoming by proposing LEMONADE, an evolutionary algorithm for
multi-objective architecture search that allows approximating the entire
Pareto-front of architectures under multiple objectives, such as predictive
performance and number of parameters, in a single run of the method. We address
the second shortcoming by proposing a Lamarckian inheritance mechanism for
LEMONADE which generates children networks that are warmstarted with the
predictive performance of their trained parents. This is accomplished by using
(approximate) network morphism operators for generating children. The
combination of these two contributions allows finding models that are on par or
even outperform both hand-crafted as well as automatically-designed networks.
|
Semi-supervised Content-based Detection of Misinformation via Tensor
Embeddings | Fake news may be intentionally created to promote economic, political and
social interests, and can lead to negative impacts on humans beliefs and
decisions. Hence, detection of fake news is an emerging problem that has become
extremely prevalent during the last few years. Most existing works on this
topic focus on manual feature extraction and supervised classification models
leveraging a large number of labeled (fake or real) articles. In contrast, we
focus on content-based detection of fake news articles, while assuming that we
have a small amount of labels, made available by manual fact-checkers or
automated sources. We argue this is a more realistic setting in the presence of
massive amounts of content, most of which cannot be easily factchecked. To that
end, we represent collections of news articles as multi-dimensional tensors,
leverage tensor decomposition to derive concise article embeddings that capture
spatial/contextual information about each news article, and use those
embeddings to create an article-by-article graph on which we propagate limited
labels. Results on three real-world datasets show that our method performs on
par or better than existing models that are fully supervised, in that we
achieve better detection accuracy using fewer labels. In particular, our
proposed method achieves 75.43% of accuracy using only 30% of labels of a
public dataset while an SVM-based classifier achieved 67.43%. Furthermore, our
method achieves 70.92% of accuracy in a large dataset using only 2% of labels.
|
Improving Native Ads CTR Prediction by Large Scale Event Embedding and
Recurrent Networks | Click through rate (CTR) prediction is very important for Native
advertisement but also hard as there is no direct query intent. In this paper
we propose a large-scale event embedding scheme to encode the each user
browsing event by training a Siamese network with weak supervision on the
users' consecutive events. The CTR prediction problem is modeled as a
supervised recurrent neural network, which naturally model the user history as
a sequence of events. Our proposed recurrent models utilizing pretrained event
embedding vectors and an attention layer to model the user history. Our
experiments demonstrate that our model significantly outperforms the baseline
and some variants.
|
Automated Detection of Adverse Drug Reactions in the Biomedical
Literature Using Convolutional Neural Networks and Biomedical Word Embeddings | Monitoring the biomedical literature for cases of Adverse Drug Reactions
(ADRs) is a critically important and time consuming task in pharmacovigilance.
The development of computer assisted approaches to aid this process in
different forms has been the subject of many recent works. One particular area
that has shown promise is the use of Deep Neural Networks, in particular,
Convolutional Neural Networks (CNNs), for the detection of ADR relevant
sentences. Using token-level convolutions and general purpose word embeddings,
this architecture has shown good performance relative to more traditional
models as well as Long Short Term Memory (LSTM) models. In this work, we
evaluate and compare two different CNN architectures using the ADE corpus. In
addition, we show that by de-duplicating the ADR relevant sentences, we can
greatly reduce overoptimism in the classification results. Finally, we evaluate
the use of word embeddings specifically developed for biomedical text and show
that they lead to a better performance in this task.
|
DOOM Level Generation using Generative Adversarial Networks | We applied Generative Adversarial Networks (GANs) to learn a model of DOOM
levels from human-designed content. Initially, we analysed the levels and
extracted several topological features. Then, for each level, we extracted a
set of images identifying the occupied area, the height map, the walls, and the
position of game objects. We trained two GANs: one using plain level images,
one using both the images and some of the features extracted during the
preliminary analysis. We used the two networks to generate new levels and
compared the results to assess whether the network trained using also the
topological features could generate levels more similar to human-designed ones.
Our results show that GANs can capture intrinsic structure of DOOM levels and
appears to be a promising approach to level generation in first person shooter
games.
|
No Metrics Are Perfect: Adversarial Reward Learning for Visual
Storytelling | Though impressive results have been achieved in visual captioning, the task
of generating abstract stories from photo streams is still a little-tapped
problem. Different from captions, stories have more expressive language styles
and contain many imaginary concepts that do not appear in the images. Thus it
poses challenges to behavioral cloning algorithms. Furthermore, due to the
limitations of automatic metrics on evaluating story quality, reinforcement
learning methods with hand-crafted rewards also face difficulties in gaining an
overall performance boost. Therefore, we propose an Adversarial REward Learning
(AREL) framework to learn an implicit reward function from human
demonstrations, and then optimize policy search with the learned reward
function. Though automatic eval- uation indicates slight performance boost over
state-of-the-art (SOTA) methods in cloning expert behaviors, human evaluation
shows that our approach achieves significant improvement in generating more
human-like stories than SOTA systems.
|
Realistic Evaluation of Deep Semi-Supervised Learning Algorithms | Semi-supervised learning (SSL) provides a powerful framework for leveraging
unlabeled data when labels are limited or expensive to obtain. SSL algorithms
based on deep neural networks have recently proven successful on standard
benchmark tasks. However, we argue that these benchmarks fail to address many
issues that these algorithms would face in real-world applications. After
creating a unified reimplementation of various widely-used SSL techniques, we
test them in a suite of experiments designed to address these issues. We find
that the performance of simple baselines which do not use unlabeled data is
often underreported, that SSL methods differ in sensitivity to the amount of
labeled and unlabeled data, and that performance can degrade substantially when
the unlabeled dataset contains out-of-class examples. To help guide SSL
research towards real-world applicability, we make our unified reimplemention
and evaluation platform publicly available.
|
On Learning Sparsely Used Dictionaries from Incomplete Samples | Most existing algorithms for dictionary learning assume that all entries of
the (high-dimensional) input data are fully observed. However, in several
practical applications (such as hyper-spectral imaging or blood glucose
monitoring), only an incomplete fraction of the data entries may be available.
For incomplete settings, no provably correct and polynomial-time algorithm has
been reported in the dictionary learning literature. In this paper, we provide
provable approaches for learning - from incomplete samples - a family of
dictionaries whose atoms have sufficiently "spread-out" mass. First, we propose
a descent-style iterative algorithm that linearly converges to the true
dictionary when provided a sufficiently coarse initial estimate. Second, we
propose an initialization algorithm that utilizes a small number of extra fully
observed samples to produce such a coarse initial estimate. Finally, we
theoretically analyze their performance and provide asymptotic statistical and
computational guarantees.
|
Semi-Supervised Learning with Declaratively Specified Entropy
Constraints | We propose a technique for declaratively specifying strategies for
semi-supervised learning (SSL). The proposed method can be used to specify
ensembles of semi-supervised learning, as well as agreement constraints and
entropic regularization constraints between these learners, and can be used to
model both well-known heuristics such as co-training and novel domain-specific
heuristics. In addition to representing individual SSL heuristics, we show that
multiple heuristics can also be automatically combined using Bayesian
optimization methods. We show consistent improvements on a suite of
well-studied SSL benchmarks, including a new state-of-the-art result on a
difficult relation extraction task.
|
DeepTriangle: A Deep Learning Approach to Loss Reserving | We propose a novel approach for loss reserving based on deep neural networks.
The approach allows for joint modeling of paid losses and claims outstanding,
and incorporation of heterogeneous inputs. We validate the models on loss
reserving data across lines of business, and show that they improve on the
predictive accuracy of existing stochastic methods. The models require minimal
feature engineering and expert input, and can be automated to produce forecasts
more frequently than manual workflows.
|
A Closer Look at Weak Label Learning for Audio Events | Audio content analysis in terms of sound events is an important research
problem for a variety of applications. Recently, the development of weak
labeling approaches for audio or sound event detection (AED) and availability
of large scale weakly labeled dataset have finally opened up the possibility of
large scale AED. However, a deeper understanding of how weak labels affect the
learning for sound events is still missing from literature. In this work, we
first describe a CNN based approach for weakly supervised training of audio
events. The approach follows some basic design principle desirable in a
learning method relying on weakly labeled audio. We then describe important
characteristics, which naturally arise in weakly supervised learning of sound
events. We show how these aspects of weak labels affect the generalization of
models. More specifically, we study how characteristics such as label density
and corruption of labels affects weakly supervised training for audio events.
We also study the feasibility of directly obtaining weak labeled data from the
web without any manual label and compare it with a dataset which has been
manually labeled. The analysis and understanding of these factors should be
taken into picture in the development of future weak label learning methods.
Audioset, a large scale weakly labeled dataset for sound events is used in our
experiments.
|
Deep Learning for Predicting Asset Returns | Deep learning searches for nonlinear factors for predicting asset returns.
Predictability is achieved via multiple layers of composite factors as opposed
to additive ones. Viewed in this way, asset pricing studies can be revisited
using multi-layer deep learners, such as rectified linear units (ReLU) or
long-short-term-memory (LSTM) for time-series effects. State-of-the-art
algorithms including stochastic gradient descent (SGD), TensorFlow and dropout
design provide imple- mentation and efficient factor exploration. To illustrate
our methodology, we revisit the equity market risk premium dataset of Welch and
Goyal (2008). We find the existence of nonlinear factors which explain
predictability of returns, in particular at the extremes of the characteristic
space. Finally, we conclude with directions for future research.
|
Where are we now? A large benchmark study of recent symbolic regression
methods | In this paper we provide a broad benchmarking of recent genetic programming
approaches to symbolic regression in the context of state of the art machine
learning approaches. We use a set of nearly 100 regression benchmark problems
culled from open source repositories across the web. We conduct a rigorous
benchmarking of four recent symbolic regression approaches as well as nine
machine learning approaches from scikit-learn. The results suggest that
symbolic regression performs strongly compared to state-of-the-art gradient
boosting algorithms, although in terms of running times is among the slowest of
the available methodologies. We discuss the results in detail and point to
future research directions that may allow symbolic regression to gain wider
adoption in the machine learning community.
|
Generalized Gaussian Kernel Adaptive Filtering | The present paper proposes generalized Gaussian kernel adaptive filtering,
where the kernel parameters are adaptive and data-driven. The Gaussian kernel
is parametrized by a center vector and a symmetric positive definite (SPD)
precision matrix, which is regarded as a generalization of the scalar width
parameter. These parameters are adaptively updated on the basis of a proposed
least-square-type rule to minimize the estimation error. The main contribution
of this paper is to establish update rules for precision matrices on the SPD
manifold in order to keep their symmetric positive-definiteness. Different from
conventional kernel adaptive filters, the proposed regressor is a superposition
of Gaussian kernels with all different parameters, which makes such regressor
more flexible. The kernel adaptive filtering algorithm is established together
with a l1-regularized least squares to avoid overfitting and the increase of
dimensionality of the dictionary. Experimental results confirm the validity of
the proposed method.
|
Driving Policy Transfer via Modularity and Abstraction | End-to-end approaches to autonomous driving have high sample complexity and
are difficult to scale to realistic urban driving. Simulation can help
end-to-end driving systems by providing a cheap, safe, and diverse training
environment. Yet training driving policies in simulation brings up the problem
of transferring such policies to the real world. We present an approach to
transferring driving policies from simulation to reality via modularity and
abstraction. Our approach is inspired by classic driving systems and aims to
combine the benefits of modular architectures and end-to-end deep learning
approaches. The key idea is to encapsulate the driving policy such that it is
not directly exposed to raw perceptual input or low-level vehicle dynamics. We
evaluate the presented approach in simulated urban environments and in the real
world. In particular, we transfer a driving policy trained in simulation to a
1/5-scale robotic truck that is deployed in a variety of conditions, with no
finetuning, on two continents. The supplementary video can be viewed at
https://youtu.be/BrMDJqI6H5U
|
Convolutional Generative Adversarial Networks with Binary Neurons for
Polyphonic Music Generation | It has been shown recently that deep convolutional generative adversarial
networks (GANs) can learn to generate music in the form of piano-rolls, which
represent music by binary-valued time-pitch matrices. However, existing models
can only generate real-valued piano-rolls and require further post-processing,
such as hard thresholding (HT) or Bernoulli sampling (BS), to obtain the final
binary-valued results. In this paper, we study whether we can have a
convolutional GAN model that directly creates binary-valued piano-rolls by
using binary neurons. Specifically, we propose to append to the generator an
additional refiner network, which uses binary neurons at the output layer. The
whole network is trained in two stages. Firstly, the generator and the
discriminator are pretrained. Then, the refiner network is trained along with
the discriminator to learn to binarize the real-valued piano-rolls the
pretrained generator creates. Experimental results show that using binary
neurons instead of HT or BS indeed leads to better results in a number of
objective measures. Moreover, deterministic binary neurons perform better than
stochastic ones in both objective measures and a subjective test. The source
code, training data and audio examples of the generated results can be found at
https://salu133445.github.io/bmusegan/ .
|
3D Consistent & Robust Segmentation of Cardiac Images by Deep Learning
with Spatial Propagation | We propose a method based on deep learning to perform cardiac segmentation on
short axis MRI image stacks iteratively from the top slice (around the base) to
the bottom slice (around the apex). At each iteration, a novel variant of U-net
is applied to propagate the segmentation of a slice to the adjacent slice below
it. In other words, the prediction of a segmentation of a slice is dependent
upon the already existing segmentation of an adjacent slice. 3D-consistency is
hence explicitly enforced. The method is trained on a large database of 3078
cases from UK Biobank. It is then tested on 756 different cases from UK Biobank
and three other state-of-the-art cohorts (ACDC with 100 cases, Sunnybrook with
30 cases, RVSC with 16 cases). Results comparable or even better than the
state-of-the-art in terms of distance measures are achieved. They also
emphasize the assets of our method, namely enhanced spatial consistency
(currently neither considered nor achieved by the state-of-the-art), and the
generalization ability to unseen cases even from other databases.
|
Generative Temporal Models with Spatial Memory for Partially Observed
Environments | In model-based reinforcement learning, generative and temporal models of
environments can be leveraged to boost agent performance, either by tuning the
agent's representations during training or via use as part of an explicit
planning mechanism. However, their application in practice has been limited to
simplistic environments, due to the difficulty of training such models in
larger, potentially partially-observed and 3D environments. In this work we
introduce a novel action-conditioned generative model of such challenging
environments. The model features a non-parametric spatial memory system in
which we store learned, disentangled representations of the environment.
Low-dimensional spatial updates are computed using a state-space model that
makes use of knowledge on the prior dynamics of the moving agent, and
high-dimensional visual observations are modelled with a Variational
Auto-Encoder. The result is a scalable architecture capable of performing
coherent predictions over hundreds of time steps across a range of partially
observed 2D and 3D environments.
|
Dynamic Few-Shot Visual Learning without Forgetting | The human visual system has the remarkably ability to be able to effortlessly
learn novel concepts from only a few examples. Mimicking the same behavior on
machine learning vision systems is an interesting and very challenging research
problem with many practical advantages on real world vision applications. In
this context, the goal of our work is to devise a few-shot visual learning
system that during test time it will be able to efficiently learn novel
categories from only a few training data while at the same time it will not
forget the initial categories on which it was trained (here called base
categories). To achieve that goal we propose (a) to extend an object
recognition system with an attention based few-shot classification weight
generator, and (b) to redesign the classifier of a ConvNet model as the cosine
similarity function between feature representations and classification weight
vectors. The latter, apart from unifying the recognition of both novel and base
categories, it also leads to feature representations that generalize better on
"unseen" categories. We extensively evaluate our approach on Mini-ImageNet
where we manage to improve the prior state-of-the-art on few-shot recognition
(i.e., we achieve 56.20% and 73.00% on the 1-shot and 5-shot settings
respectively) while at the same time we do not sacrifice any accuracy on the
base categories, which is a characteristic that most prior approaches lack.
Finally, we apply our approach on the recently introduced few-shot benchmark of
Bharath and Girshick [4] where we also achieve state-of-the-art results. The
code and models of our paper will be published on:
https://github.com/gidariss/FewShotWithoutForgetting
|
Structured Pruning for Efficient ConvNets via Incremental Regularization | Parameter pruning is a promising approach for CNN compression and
acceleration by eliminating redundant model parameters with tolerable
performance degrade. Despite its effectiveness, existing regularization-based
parameter pruning methods usually drive weights towards zero with large and
constant regularization factors, which neglects the fragility of the
expressiveness of CNNs, and thus calls for a more gentle regularization scheme
so that the networks can adapt during pruning. To achieve this, we propose a
new and novel regularization-based pruning method, named IncReg, to
incrementally assign different regularization factors to different weights
based on their relative importance. Empirical analysis on CIFAR-10 dataset
verifies the merits of IncReg. Further extensive experiments with popular CNNs
on CIFAR-10 and ImageNet datasets show that IncReg achieves comparable to even
better results compared with state-of-the-arts. Our source codes and trained
models are available here: https://github.com/mingsun-tse/caffe_increg.
|
Estimation with Low-Rank Time-Frequency Synthesis Models | Many state-of-the-art signal decomposition techniques rely on a low-rank
factorization of a time-frequency (t-f) transform. In particular, nonnegative
matrix factorization (NMF) of the spectrogram has been considered in many audio
applications. This is an analysis approach in the sense that the factorization
is applied to the squared magnitude of the analysis coefficients returned by
the t-f transform. In this paper we instead propose a synthesis approach, where
low-rankness is imposed to the synthesis coefficients of the data signal over a
given t-f dictionary (such as a Gabor frame). As such we offer a novel modeling
paradigm that bridges t-f synthesis modeling and traditional analysis-based NMF
approaches. The proposed generative model allows in turn to design more
sophisticated multi-layer representations that can efficiently capture diverse
forms of structure. Additionally, the generative modeling allows to exploit t-f
low-rankness for compressive sensing. We present efficient iterative shrinkage
algorithms to perform estimation in the proposed models and illustrate the
capabilities of the new modeling paradigm over audio signal processing
examples.
|
Unsupervised Disentangled Representation Learning with Analogical
Relations | Learning the disentangled representation of interpretable generative factors
of data is one of the foundations to allow artificial intelligence to think
like people. In this paper, we propose the analogical training strategy for the
unsupervised disentangled representation learning in generative models. The
analogy is one of the typical cognitive processes, and our proposed strategy is
based on the observation that sample pairs in which one is different from the
other in one specific generative factor show the same analogical relation.
Thus, the generator is trained to generate sample pairs from which a designed
classifier can identify the underlying analogical relation. In addition, we
propose a disentanglement metric called the subspace score, which is inspired
by subspace learning methods and does not require supervised information.
Experiments show that our proposed training strategy allows the generative
models to find the disentangled factors, and that our methods can give
competitive performances as compared with the state-of-the-art methods.
|
Strong Baselines for Neural Semi-supervised Learning under Domain Shift | Novel neural models have been proposed in recent years for learning under
domain shift. Most models, however, only evaluate on a single task, on
proprietary datasets, or compare to weak baselines, which makes comparison of
models difficult. In this paper, we re-evaluate classic general-purpose
bootstrapping approaches in the context of neural networks under domain shifts
vs. recent neural approaches and propose a novel multi-task tri-training method
that reduces the time and space complexity of classic tri-training. Extensive
experiments on two benchmarks are negative: while our novel method establishes
a new state-of-the-art for sentiment analysis, it does not fare consistently
the best. More importantly, we arrive at the somewhat surprising conclusion
that classic tri-training, with some additions, outperforms the state of the
art. We conclude that classic approaches constitute an important and strong
baseline.
|
Hand Pose Estimation via Latent 2.5D Heatmap Regression | Estimating the 3D pose of a hand is an essential part of human-computer
interaction. Estimating 3D pose using depth or multi-view sensors has become
easier with recent advances in computer vision, however, regressing pose from a
single RGB image is much less straightforward. The main difficulty arises from
the fact that 3D pose requires some form of depth estimates, which are
ambiguous given only an RGB image. In this paper we propose a new method for 3D
hand pose estimation from a monocular image through a novel 2.5D pose
representation. Our new representation estimates pose up to a scaling factor,
which can be estimated additionally if a prior of the hand size is given. We
implicitly learn depth maps and heatmap distributions with a novel CNN
architecture. Our system achieves the state-of-the-art estimation of 2D and 3D
hand pose on several challenging datasets in presence of severe occlusions.
|
QANet: Combining Local Convolution with Global Self-Attention for
Reading Comprehension | Current end-to-end machine reading and question answering (Q\&A) models are
primarily based on recurrent neural networks (RNNs) with attention. Despite
their success, these models are often slow for both training and inference due
to the sequential nature of RNNs. We propose a new Q\&A architecture called
QANet, which does not require recurrent networks: Its encoder consists
exclusively of convolution and self-attention, where convolution models local
interactions and self-attention models global interactions. On the SQuAD
dataset, our model is 3x to 13x faster in training and 4x to 9x faster in
inference, while achieving equivalent accuracy to recurrent models. The
speed-up gain allows us to train the model with much more data. We hence
combine our model with data generated by backtranslation from a neural machine
translation model. On the SQuAD dataset, our single model, trained with
augmented data, achieves 84.6 F1 score on the test set, which is significantly
better than the best published F1 score of 81.8.
|
Stochastic Conditional Gradient Methods: From Convex Minimization to
Submodular Maximization | This paper considers stochastic optimization problems for a large class of
objective functions, including convex and continuous submodular. Stochastic
proximal gradient methods have been widely used to solve such problems;
however, their applicability remains limited when the problem dimension is
large and the projection onto a convex set is costly. Instead, stochastic
conditional gradient methods are proposed as an alternative solution relying on
(i) Approximating gradients via a simple averaging technique requiring a single
stochastic gradient evaluation per iteration; (ii) Solving a linear program to
compute the descent/ascent direction. The averaging technique reduces the noise
of gradient approximations as time progresses, and replacing projection step in
proximal methods by a linear program lowers the computational complexity of
each iteration. We show that under convexity and smoothness assumptions, our
proposed method converges to the optimal objective function value at a
sublinear rate of $O(1/t^{1/3})$. Further, for a monotone and continuous
DR-submodular function and subject to a general convex body constraint, we
prove that our proposed method achieves a $((1-1/e)OPT-\eps)$ guarantee with
$O(1/\eps^3)$ stochastic gradient computations. This guarantee matches the
known hardness results and closes the gap between deterministic and stochastic
continuous submodular maximization. Additionally, we obtain $((1/e)OPT -\eps)$
guarantee after using $O(1/\eps^3)$ stochastic gradients for the case that the
objective function is continuous DR-submodular but non-monotone and the
constraint set is down-closed. By using stochastic continuous optimization as
an interface, we provide the first $(1-1/e)$ tight approximation guarantee for
maximizing a monotone but stochastic submodular set function subject to a
matroid constraint and $(1/e)$ approximation guarantee for the non-monotone
case.
|
A Visual Distance for WordNet | Measuring the distance between concepts is an important field of study of
Natural Language Processing, as it can be used to improve tasks related to the
interpretation of those same concepts. WordNet, which includes a wide variety
of concepts associated with words (i.e., synsets), is often used as a source
for computing those distances. In this paper, we explore a distance for WordNet
synsets based on visual features, instead of lexical ones. For this purpose, we
extract the graphic features generated within a deep convolutional neural
networks trained with ImageNet and use those features to generate a
representative of each synset. Based on those representatives, we define a
distance measure of synsets, which complements the traditional lexical
distances. Finally, we propose some experiments to evaluate its performance and
compare it with the current state-of-the-art.
|
On The Complexity of Sparse Label Propagation | This paper investigates the computational complexity of sparse label
propagation which has been proposed recently for processing network structured
data. Sparse label propagation amounts to a convex optimization problem and
might be considered as an extension of basis pursuit from sparse vectors to
network structured datasets. Using a standard first-order oracle model, we
characterize the number of iterations for sparse label propagation to achieve a
prescribed accuracy. In particular, we derive an upper bound on the number of
iterations required to achieve a certain accuracy and show that this upper
bound is sharp for datasets having a chain structure (e.g., time series).
|
A comparative study of feature selection methods for stress hotspot
classification in materials | The first step in constructing a machine learning model is defining the
features of the data set that can be used for optimal learning. In this work we
discuss feature selection methods, which can be used to build better models, as
well as achieve model interpretability. We applied these methods in the context
of stress hotspot classification problem, to determine what microstructural
characteristics can cause stress to build up in certain grains during uniaxial
tensile deformation. The results show how some feature selection techniques are
biased and demonstrate a preferred technique to get feature rankings for
physical interpretations.
|
When is there a Representer Theorem? Nondifferentiable Regularisers and
Banach spaces | We consider a general regularised interpolation problem for learning a
parameter vector from data. The well known representer theorem says that under
certain conditions on the regulariser there exists a solution in the linear
span of the data points. This is the core of kernel methods in machine learning
as it makes the problem computationally tractable. Necessary and sufficient
conditions for differentiable regularisers on Hilbert spaces to admit a
representer theorem have been proved. We extend those results to
nondifferentiable regularisers on uniformly convex and uniformly smooth Banach
spaces. This gives a (more) complete answer to the question when there is a
representer theorem. We then note that for regularised interpolation in fact
the solution is determined by the function space alone and independent of the
regulariser, making the extension to Banach spaces even more valuable.
|
Identifying and Alleviating Concept Drift in Streaming Tensor
Decomposition | Tensor decompositions are used in various data mining applications from
social network to medical applications and are extremely useful in discovering
latent structures or concepts in the data. Many real-world applications are
dynamic in nature and so are their data. To deal with this dynamic nature of
data, there exist a variety of online tensor decomposition algorithms. A
central assumption in all those algorithms is that the number of latent
concepts remains fixed throughout the entire stream. However, this need not be
the case. Every incoming batch in the stream may have a different number of
latent concepts, and the difference in latent concepts from one tensor batch to
another can provide insights into how our findings in a particular application
behave and deviate over time. In this paper, we define "concept" and "concept
drift" in the context of streaming tensor decomposition, as the manifestation
of the variability of latent concepts throughout the stream. Furthermore, we
introduce SeekAndDestroy, an algorithm that detects concept drift in streaming
tensor decomposition and is able to produce results robust to that drift. To
the best of our knowledge, this is the first work that investigates concept
drift in streaming tensor decomposition. We extensively evaluate SeekAndDestroy
on synthetic datasets, which exhibit a wide variety of realistic drift. Our
experiments demonstrate the effectiveness of SeekAndDestroy, both in the
detection of concept drift and in the alleviation of its effects, producing
results with similar quality to decomposing the entire tensor in one shot.
Additionally, in real datasets, SeekAndDestroy outperforms other streaming
baselines, while discovering novel useful components.
|
Convergence guarantees for a class of non-convex and non-smooth
optimization problems | We consider the problem of finding critical points of functions that are
non-convex and non-smooth. Studying a fairly broad class of such problems, we
analyze the behavior of three gradient-based methods (gradient descent,
proximal update, and Frank-Wolfe update). For each of these methods, we
establish rates of convergence for general problems, and also prove faster
rates for continuous sub-analytic functions. We also show that our algorithms
can escape strict saddle points for a class of non-smooth functions, thereby
generalizing known results for smooth functions. Our analysis leads to a
simplification of the popular CCCP algorithm, used for optimizing functions
that can be written as a difference of two convex functions. Our simplified
algorithm retains all the convergence properties of CCCP, along with a
significantly lower cost per iteration. We illustrate our methods and theory
via applications to the problems of best subset selection, robust estimation,
mixture density estimation, and shape-from-shading reconstruction.
|
Towards Fast Computation of Certified Robustness for ReLU Networks | Verifying the robustness property of a general Rectified Linear Unit (ReLU)
network is an NP-complete problem [Katz, Barrett, Dill, Julian and Kochenderfer
CAV17]. Although finding the exact minimum adversarial distortion is hard,
giving a certified lower bound of the minimum distortion is possible. Current
available methods of computing such a bound are either time-consuming or
delivering low quality bounds that are too loose to be useful. In this paper,
we exploit the special structure of ReLU networks and provide two
computationally efficient algorithms Fast-Lin and Fast-Lip that are able to
certify non-trivial lower bounds of minimum distortions, by bounding the ReLU
units with appropriate linear functions Fast-Lin, or by bounding the local
Lipschitz constant Fast-Lip. Experiments show that (1) our proposed methods
deliver bounds close to (the gap is 2-3X) exact minimum distortion found by
Reluplex in small MNIST networks while our algorithms are more than 10,000
times faster; (2) our methods deliver similar quality of bounds (the gap is
within 35% and usually around 10%; sometimes our bounds are even better) for
larger networks compared to the methods based on solving linear programming
problems but our algorithms are 33-14,000 times faster; (3) our method is
capable of solving large MNIST and CIFAR networks up to 7 layers with more than
10,000 neurons within tens of seconds on a single CPU core.
In addition, we show that, in fact, there is no polynomial time algorithm
that can approximately find the minimum $\ell_1$ adversarial distortion of a
ReLU network with a $0.99\ln n$ approximation ratio unless
$\mathsf{NP}$=$\mathsf{P}$, where $n$ is the number of neurons in the network.
|
End-to-End Multimodal Speech Recognition | Transcription or sub-titling of open-domain videos is still a challenging
domain for Automatic Speech Recognition (ASR) due to the data's challenging
acoustics, variable signal processing and the essentially unrestricted domain
of the data. In previous work, we have shown that the visual channel --
specifically object and scene features -- can help to adapt the acoustic model
(AM) and language model (LM) of a recognizer, and we are now expanding this
work to end-to-end approaches. In the case of a Connectionist Temporal
Classification (CTC)-based approach, we retain the separation of AM and LM,
while for a sequence-to-sequence (S2S) approach, both information sources are
adapted together, in a single model. This paper also analyzes the behavior of
CTC and S2S models on noisy video data (How-To corpus), and compares it to
results on the clean Wall Street Journal (WSJ) corpus, providing insight into
the robustness of both approaches.
|
RULLS: Randomized Union of Locally Linear Subspaces for Feature
Engineering | Feature engineering plays an important role in the success of a machine
learning model. Most of the effort in training a model goes into data
preparation and choosing the right representation. In this paper, we propose a
robust feature engineering method, Randomized Union of Locally Linear Subspaces
(RULLS). We generate sparse, non-negative, and rotation invariant features in
an unsupervised fashion. RULLS aggregates features from a random union of
subspaces by describing each point using globally chosen landmarks. These
landmarks serve as anchor points for choosing subspaces. Our method provides a
way to select features that are relevant in the neighborhood around these
chosen landmarks. Distances from each data point to $k$ closest landmarks are
encoded in the feature matrix. The final feature representation is a union of
features from all chosen subspaces.
The effectiveness of our algorithm is shown on various real-world datasets
for tasks such as clustering and classification of raw data and in the presence
of noise. We compare our method with existing feature generation methods.
Results show a high performance of our method on both classification and
clustering tasks.
|
Mathematical Analysis on Out-of-Sample Extensions | Let $X=\mathbf{X}\cup\mathbf{Z}$ be a data set in $\mathbb{R}^D$, where
$\mathbf{X}$ is the training set and $\mathbf{Z}$ is the test one. Many
unsupervised learning algorithms based on kernel methods have been developed to
provide dimensionality reduction (DR) embedding for a given training set $\Phi:
\mathbf{X} \to \mathbb{R}^d$ ( $d\ll D$) that maps the high-dimensional data
$\mathbf{X}$ to its low-dimensional feature representation
$\mathbf{Y}=\Phi(\mathbf{X})$. However, these algorithms do not
straightforwardly produce DR of the test set $\mathbf{Z}$. An out-of-sample
extension method provides DR of $\mathbf{Z}$ using an extension of the existent
embedding $\Phi$, instead of re-computing the DR embedding for the whole set
$X$. Among various out-of-sample DR extension methods, those based on
Nystr\"{o}m approximation are very attractive. Many papers have developed such
out-of-extension algorithms and shown their validity by numerical experiments.
However, the mathematical theory for the DR extension still need further
consideration. Utilizing the reproducing kernel Hilbert space (RKHS) theory,
this paper develops a preliminary mathematical analysis on the out-of-sample DR
extension operators. It treats an out-of-sample DR extension operator as an
extension of the identity on the RKHS defined on $\mathbf{X}$. Then the
Nystr\"{o}m-type DR extension turns out to be an orthogonal projection. In the
paper, we also present the conditions for the exact DR extension and give the
estimate for the error of the extension.
|
Multi-Layer Sparse Coding: The Holistic Way | The recently proposed multi-layer sparse model has raised insightful
connections between sparse representations and convolutional neural networks
(CNN). In its original conception, this model was restricted to a cascade of
convolutional synthesis representations. In this paper, we start by addressing
a more general model, revealing interesting ties to fully connected networks.
We then show that this multi-layer construction admits a brand new
interpretation in a unique symbiosis between synthesis and analysis models:
while the deepest layer indeed provides a synthesis representation, the
mid-layers decompositions provide an analysis counterpart. This new perspective
exposes the suboptimality of previously proposed pursuit approaches, as they do
not fully leverage all the information comprised in the model constraints.
Armed with this understanding, we address fundamental theoretical issues,
revisiting previous analysis and expanding it. Motivated by the limitations of
previous algorithms, we then propose an integrated - holistic - alternative
that estimates all representations in the model simultaneously, and analyze all
these different schemes under stochastic noise assumptions. Inspired by the
synthesis-analysis duality, we further present a Holistic Pursuit algorithm,
which alternates between synthesis and analysis sparse coding steps, eventually
solving for the entire model as a whole, with provable improved performance.
Finally, we present numerical results that demonstrate the practical advantages
of our approach.
|
Off the Beaten Track: Using Deep Learning to Interpolate Between Music
Genres | We describe a system based on deep learning that generates drum patterns in
the electronic dance music domain. Experimental results reveal that generated
patterns can be employed to produce musically sound and creative transitions
between different genres, and that the process of generation is of interest to
practitioners in the field.
|
Improved Classification Based on Deep Belief Networks | For better classification generative models are used to initialize the model
and model features before training a classifier. Typically it is needed to
solve separate unsupervised and supervised learning problems. Generative
restricted Boltzmann machines and deep belief networks are widely used for
unsupervised learning. We developed several supervised models based on DBN in
order to improve this two-phase strategy. Modifying the loss function to
account for expectation with respect to the underlying generative model,
introducing weight bounds, and multi-level programming are applied in model
development. The proposed models capture both unsupervised and supervised
objectives effectively. The computational study verifies that our models
perform better than the two-phase training approach.
|
HG-means: A scalable hybrid genetic algorithm for minimum sum-of-squares
clustering | Minimum sum-of-squares clustering (MSSC) is a widely used clustering model,
of which the popular K-means algorithm constitutes a local minimizer. It is
well known that the solutions of K-means can be arbitrarily distant from the
true MSSC global optimum, and dozens of alternative heuristics have been
proposed for this problem. However, no other algorithm has been predominantly
adopted in the literature. This may be related to differences of computational
effort, or to the assumption that a near-optimal solution of the MSSC has only
a marginal impact on clustering validity. In this article, we dispute this
belief. We introduce an efficient population-based metaheuristic that uses
K-means as a local search in combination with problem-tailored crossover,
mutation, and diversification operators. This algorithm can be interpreted as a
multi-start K-means, in which the initial center positions are carefully
sampled based on the search history. The approach is scalable and accurate,
outperforming all recent state-of-the-art algorithms for MSSC in terms of
solution quality, measured by the depth of local minima. This enhanced accuracy
leads to clusters which are significantly closer to the ground truth than those
of other algorithms, for overlapping Gaussian-mixture datasets with a large
number of features. Therefore, improved global optimization methods appear to
be essential to better exploit the MSSC model in high dimension.
|
On the Dual Geometry of Laplacian Eigenfunctions | We discuss the geometry of Laplacian eigenfunctions $-\Delta \phi = \lambda
\phi$ on compact manifolds $(M,g)$ and combinatorial graphs $G=(V,E)$. The
'dual' geometry of Laplacian eigenfunctions is well understood on
$\mathbb{T}^d$ (identified with $\mathbb{Z}^d$) and $\mathbb{R}^n$ (which is
self-dual). The dual geometry is of tremendous role in various fields of pure
and applied mathematics. The purpose of our paper is to point out a notion of
similarity between eigenfunctions that allows to reconstruct that geometry. Our
measure of 'similarity' $ \alpha(\phi_{\lambda}, \phi_{\mu})$ between
eigenfunctions $\phi_{\lambda}$ and $\phi_{\mu}$ is given by a global average
of local correlations $$ \alpha(\phi_{\lambda}, \phi_{\mu})^2 = \|
\phi_{\lambda} \phi_{\mu} \|_{L^2}^{-2}\int_{M}{ \left( \int_{M}{ p(t,x,y)(
\phi_{\lambda}(y) - \phi_{\lambda}(x))( \phi_{\mu}(y) - \phi_{\mu}(x)) dy}
\right)^2 dx},$$ where $p(t,x,y)$ is the classical heat kernel and $e^{-t
\lambda} + e^{-t \mu} = 1$. This notion recovers all classical notions of
duality but is equally applicable to other (rough) geometries and graphs; many
numerical examples in different continuous and discrete settings illustrate the
result.
|
Hierarchical Density Order Embeddings | By representing words with probability densities rather than point vectors,
probabilistic word embeddings can capture rich and interpretable semantic
information and uncertainty. The uncertainty information can be particularly
meaningful in capturing entailment relationships -- whereby general words such
as "entity" correspond to broad distributions that encompass more specific
words such as "animal" or "instrument". We introduce density order embeddings,
which learn hierarchical representations through encapsulation of probability
densities. In particular, we propose simple yet effective loss functions and
distance metrics, as well as graph-based schemes to select negative samples to
better learn hierarchical density representations. Our approach provides
state-of-the-art performance on the WordNet hypernym relationship prediction
task and the challenging HyperLex lexical entailment dataset -- while retaining
a rich and interpretable density representation.
|
Generative Model for Heterogeneous Inference | Generative models (GMs) such as Generative Adversary Network (GAN) and
Variational Auto-Encoder (VAE) have thrived these years and achieved high
quality results in generating new samples. Especially in Computer Vision, GMs
have been used in image inpainting, denoising and completion, which can be
treated as the inference from observed pixels to corrupted pixels. However,
images are hierarchically structured which are quite different from many
real-world inference scenarios with non-hierarchical features. These inference
scenarios contain heterogeneous stochastic variables and irregular mutual
dependences. Traditionally they are modeled by Bayesian Network (BN). However,
the learning and inference of BN model are NP-hard thus the number of
stochastic variables in BN is highly constrained. In this paper, we adapt
typical GMs to enable heterogeneous learning and inference in polynomial
time.We also propose an extended autoregressive (EAR) model and an EAR with
adversary loss (EARA) model and give theoretical results on their
effectiveness. Experiments on several BN datasets show that our proposed EAR
model achieves the best performance in most cases compared to other GMs. Except
for black box analysis, we've also done a serial of experiments on Markov
border inference of GMs for white box analysis and give theoretical results.
|
Competitive Learning Enriches Learning Representation and Accelerates
the Fine-tuning of CNNs | In this study, we propose the integration of competitive learning into
convolutional neural networks (CNNs) to improve the representation learning and
efficiency of fine-tuning. Conventional CNNs use back propagation learning, and
it enables powerful representation learning by a discrimination task. However,
it requires huge amount of labeled data, and acquisition of labeled data is
much harder than that of unlabeled data. Thus, efficient use of unlabeled data
is getting crucial for DNNs. To address the problem, we introduce unsupervised
competitive learning into the convolutional layer, and utilize unlabeled data
for effective representation learning. The results of validation experiments
using a toy model demonstrated that strong representation learning effectively
extracted bases of images into convolutional filters using unlabeled data, and
accelerated the speed of the fine-tuning of subsequent supervised back
propagation learning. The leverage was more apparent when the number of filters
was sufficiently large, and, in such a case, the error rate steeply decreased
in the initial phase of fine-tuning. Thus, the proposed method enlarged the
number of filters in CNNs, and enabled a more detailed and generalized
representation. It could provide a possibility of not only deep but broad
neural networks.
|
A Neural Embeddings Approach for Detecting Mobile Counterfeit Apps | Counterfeit apps impersonate existing popular apps in attempts to misguide
users to install them for various reasons such as collecting personal
information, spreading malware, or simply to increase their advertisement
revenue. Many counterfeits can be identified once installed, however even a
tech-savvy user may struggle to detect them before installation as app icons
and descriptions can be quite similar to the original app. To this end, this
paper proposes to use neural embeddings generated by state-of-the-art
convolutional neural networks (CNNs) to measure the similarity between images.
Our results show that for the problem of counterfeit detection a novel approach
of using style embeddings given by the Gram matrix of CNN filter responses
outperforms baseline methods such as content embeddings and SIFT features. We
show that further performance increases can be achieved by combining style
embeddings with content embeddings. We present an analysis of approximately 1.2
million apps from Google Play Store and identify a set of potential
counterfeits for top-1,000 apps. Under a conservative assumption, we were able
to find 139 apps that contain malware in a set of 6,880 apps that showed high
visual similarity to one of the top-1,000 apps in Google Play Store.
|
Random Fourier Features for Kernel Ridge Regression: Approximation
Bounds and Statistical Guarantees | Random Fourier features is one of the most popular techniques for scaling up
kernel methods, such as kernel ridge regression. However, despite impressive
empirical results, the statistical properties of random Fourier features are
still not well understood. In this paper we take steps toward filling this gap.
Specifically, we approach random Fourier features from a spectral matrix
approximation point of view, give tight bounds on the number of Fourier
features required to achieve a spectral approximation, and show how spectral
matrix approximation bounds imply statistical guarantees for kernel ridge
regression.
Qualitatively, our results are twofold: on the one hand, we show that random
Fourier feature approximation can provably speed up kernel ridge regression
under reasonable assumptions. At the same time, we show that the method is
suboptimal, and sampling from a modified distribution in Fourier space, given
by the leverage function of the kernel, yields provably better performance. We
study this optimal sampling distribution for the Gaussian kernel, achieving a
nearly complete characterization for the case of low-dimensional bounded
datasets. Based on this characterization, we propose an efficient sampling
scheme with guarantees superior to random Fourier features in this regime.
|
High-dimensional Penalty Selection via Minimum Description Length
Principle | We tackle the problem of penalty selection of regularization on the basis of
the minimum description length (MDL) principle. In particular, we consider that
the design space of the penalty function is high-dimensional. In this
situation, the luckiness-normalized-maximum-likelihood(LNML)-minimization
approach is favorable, because LNML quantifies the goodness of regularized
models with any forms of penalty functions in view of the minimum description
length principle, and guides us to a good penalty function through the
high-dimensional space. However, the minimization of LNML entails two major
challenges: 1) the computation of the normalizing factor of LNML and 2) its
minimization in high-dimensional spaces. In this paper, we present a novel
regularization selection method (MDL-RS), in which a tight upper bound of LNML
(uLNML) is minimized with local convergence guarantee. Our main contribution is
the derivation of uLNML, which is a uniform-gap upper bound of LNML in an
analytic expression. This solves the above challenges in an approximate manner
because it allows us to accurately approximate LNML and then efficiently
minimize it. The experimental results show that MDL-RS improves the
generalization performance of regularized estimates specifically when the model
has redundant parameters.
|
Profile-guided memory optimization for deep neural networks | Recent years have seen deep neural networks (DNNs) becoming wider and deeper
to achieve better performance in many applications of AI. Such DNNs however
require huge amounts of memory to store weights and intermediate results (e.g.,
activations, feature maps, etc.) in propagation. This requirement makes it
difficult to run the DNNs on devices with limited, hard-to-extend memory,
degrades the running time performance, and restricts the design of network
models. We address this challenge by developing a novel profile-guided memory
optimization to efficiently and quickly allocate memory blocks during the
propagation in DNNs. The optimization utilizes a simple and fast heuristic
algorithm based on the two-dimensional rectangle packing problem. Experimenting
with well-known neural network models, we confirm that our method not only
reduces the memory consumption by up to $49.5\%$ but also accelerates training
and inference by up to a factor of four thanks to the rapidity of the memory
allocation and the ability to use larger mini-batch sizes.
|
Candidate Labeling for Crowd Learning | Crowdsourcing has become very popular among the machine learning community as
a way to obtain labels that allow a ground truth to be estimated for a given
dataset. In most of the approaches that use crowdsourced labels, annotators are
asked to provide, for each presented instance, a single class label. Such a
request could be inefficient, that is, considering that the labelers may not be
experts, that way to proceed could fail to take real advantage of the knowledge
of the labelers. In this paper, the use of candidate labeling for crowd
learning is proposed, where the annotators may provide more than a single label
per instance to try not to miss the real label. The main hypothesis is that, by
allowing candidate labeling, knowledge can be extracted from the labelers more
efficiently by than in the standard crowd learning scenario. Empirical evidence
which supports that hypothesis is presented.
|
Extended Vertical Lists for Temporal Pattern Mining from Multivariate
Time Series | Temporal Pattern Mining (TPM) is the problem of mining predictive complex
temporal patterns from multivariate time series in a supervised setting. We
develop a new method called the Fast Temporal Pattern Mining with Extended
Vertical Lists. This method utilizes an extension of the Apriori property which
requires a more complex pattern to appear within records only at places where
all of its subpatterns are detected as well. The approach is based on a novel
data structure called the Extended Vertical List that tracks positions of the
first state of the pattern inside records. Extensive computational results
indicate that the new method performs significantly faster than the previous
version of the algorithm for TMP. However, the speed-up comes at the expense of
memory usage.
|
Decentralized learning with budgeted network load using Gaussian copulas
and classifier ensembles | We examine a network of learners which address the same classification task
but must learn from different data sets. The learners cannot share data but
instead share their models. Models are shared only one time so as to preserve
the network load. We introduce DELCO (standing for Decentralized Ensemble
Learning with COpulas), a new approach allowing to aggregate the predictions of
the classifiers trained by each learner. The proposed method aggregates the
base classifiers using a probabilistic model relying on Gaussian copulas.
Experiments on logistic regressor ensembles demonstrate competing accuracy and
increased robustness in case of dependent classifiers. A companion python
implementation can be downloaded at https://github.com/john-klein/DELCO
|
Adaptive pooling operators for weakly labeled sound event detection | Sound event detection (SED) methods are tasked with labeling segments of
audio recordings by the presence of active sound sources. SED is typically
posed as a supervised machine learning problem, requiring strong annotations
for the presence or absence of each sound source at every time instant within
the recording. However, strong annotations of this type are both labor- and
cost-intensive for human annotators to produce, which limits the practical
scalability of SED methods.
In this work, we treat SED as a multiple instance learning (MIL) problem,
where training labels are static over a short excerpt, indicating the presence
or absence of sound sources but not their temporal locality. The models,
however, must still produce temporally dynamic predictions, which must be
aggregated (pooled) when comparing against static labels during training. To
facilitate this aggregation, we develop a family of adaptive pooling
operators---referred to as auto-pool---which smoothly interpolate between
common pooling operators, such as min-, max-, or average-pooling, and
automatically adapt to the characteristics of the sound sources in question. We
evaluate the proposed pooling operators on three datasets, and demonstrate that
in each case, the proposed methods outperform non-adaptive pooling operators
for static prediction, and nearly match the performance of models trained with
strong, dynamic annotations. The proposed method is evaluated in conjunction
with convolutional neural networks, but can be readily applied to any
differentiable model for time-series label prediction.
|
Notes on stable learning with piecewise-linear basis functions | We discuss technical results on learning function approximations using
piecewise-linear basis functions, and analyze their stability and convergence
using nonlinear contraction theory.
|
Quantized Compressive K-Means | The recent framework of compressive statistical learning aims at designing
tractable learning algorithms that use only a heavily compressed
representation-or sketch-of massive datasets. Compressive K-Means (CKM) is such
a method: it estimates the centroids of data clusters from pooled, non-linear,
random signatures of the learning examples. While this approach significantly
reduces computational time on very large datasets, its digital implementation
wastes acquisition resources because the learning examples are compressed only
after the sensing stage. The present work generalizes the sketching procedure
initially defined in Compressive K-Means to a large class of periodic
nonlinearities including hardware-friendly implementations that compressively
acquire entire datasets. This idea is exemplified in a Quantized Compressive
K-Means procedure, a variant of CKM that leverages 1-bit universal quantization
(i.e. retaining the least significant bit of a standard uniform quantizer) as
the periodic sketch nonlinearity. Trading for this resource-efficient signature
(standard in most acquisition schemes) has almost no impact on the clustering
performances, as illustrated by numerical experiments.
|
Securing Distributed Gradient Descent in High Dimensional Statistical
Learning | We consider unreliable distributed learning systems wherein the training data
is kept confidential by external workers, and the learner has to interact
closely with those workers to train a model. In particular, we assume that
there exists a system adversary that can adaptively compromise some workers;
the compromised workers deviate from their local designed specifications by
sending out arbitrarily malicious messages.
We assume in each communication round, up to $q$ out of the $m$ workers
suffer Byzantine faults. Each worker keeps a local sample of size $n$ and the
total sample size is $N=nm$. We propose a secured variant of the gradient
descent method that can tolerate up to a constant fraction of Byzantine
workers, i.e., $q/m = O(1)$. Moreover, we show the statistical estimation error
of the iterates converges in $O(\log N)$ rounds to $O(\sqrt{q/N} +
\sqrt{d/N})$, where $d$ is the model dimension. As long as $q=O(d)$, our
proposed algorithm achieves the optimal error rate $O(\sqrt{d/N})$. Our results
are obtained under some technical assumptions. Specifically, we assume
strongly-convex population risk. Nevertheless, the empirical risk (sample
version) is allowed to be non-convex. The core of our method is to robustly
aggregate the gradients computed by the workers based on the filtering
procedure proposed by Steinhardt et al. On the technical front, deviating from
the existing literature on robustly estimating a finite-dimensional mean
vector, we establish a {\em uniform} concentration of the sample covariance
matrix of gradients, and show that the aggregated gradient, as a function of
model parameter, converges uniformly to the true gradient function. To get a
near-optimal uniform concentration bound, we develop a new matrix concentration
inequality, which might be of independent interest.
|
BEST : A decision tree algorithm that handles missing values | The main contribution of this paper is the development of a new decision tree
algorithm. The proposed approach allows users to guide the algorithm through
the data partitioning process. We believe this feature has many applications
but in this paper we demonstrate how to utilize this algorithm to analyse data
sets containing missing values. We tested our algorithm against simulated data
sets with various missing data structures and a real data set. The results
demonstrate that this new classification procedure efficiently handles missing
values and produces results that are slightly more accurate and more
interpretable than most common procedures without any imputations or
pre-processing.
|
Modeling Psychotherapy Dialogues with Kernelized Hashcode
Representations: A Nonparametric Information-Theoretic Approach | We propose a novel dialogue modeling framework, the first-ever nonparametric
kernel functions based approach for dialogue modeling, which learns kernelized
hashcodes as compressed text representations; unlike traditional deep learning
models, it handles well relatively small datasets, while also scaling to large
ones. We also derive a novel lower bound on mutual information, used as a
model-selection criterion favoring representations with better alignment
between the utterances of participants in a collaborative dialogue setting, as
well as higher predictability of the generated responses. As demonstrated on
three real-life datasets, including prominently psychotherapy sessions, the
proposed approach significantly outperforms several state-of-art neural network
based dialogue systems, both in terms of computational efficiency, reducing
training time from days or weeks to hours, and the response quality, achieving
an order of magnitude improvement over competitors in frequency of being chosen
as the best model by human evaluators.
|
The loss landscape of overparameterized neural networks | We explore some mathematical features of the loss landscape of
overparameterized neural networks. A priori one might imagine that the loss
function looks like a typical function from $\mathbb{R}^n$ to $\mathbb{R}$ - in
particular, nonconvex, with discrete global minima. In this paper, we prove
that in at least one important way, the loss function of an overparameterized
neural network does not look like a typical function. If a neural net has $n$
parameters and is trained on $d$ data points, with $n>d$, we show that the
locus $M$ of global minima of $L$ is usually not discrete, but rather an $n-d$
dimensional submanifold of $\mathbb{R}^n$. In practice, neural nets commonly
have orders of magnitude more parameters than data points, so this observation
implies that $M$ is typically a very high-dimensional subset of $\mathbb{R}^n$.
|
End-to-End Speech Separation with Unfolded Iterative Phase
Reconstruction | This paper proposes an end-to-end approach for single-channel
speaker-independent multi-speaker speech separation, where time-frequency (T-F)
masking, the short-time Fourier transform (STFT), and its inverse are
represented as layers within a deep network. Previous approaches, rather than
computing a loss on the reconstructed signal, used a surrogate loss based on
the target STFT magnitudes. This ignores reconstruction error introduced by
phase inconsistency. In our approach, the loss function is directly defined on
the reconstructed signals, which are optimized for best separation. In
addition, we train through unfolded iterations of a phase reconstruction
algorithm, represented as a series of STFT and inverse STFT layers. While mask
values are typically limited to lie between zero and one for approaches using
the mixture phase for reconstruction, this limitation is less relevant if the
estimated magnitudes are to be used together with phase reconstruction. We thus
propose several novel activation functions for the output layer of the T-F
masking, to allow mask values beyond one. On the publicly-available wsj0-2mix
dataset, our approach achieves state-of-the-art 12.6 dB scale-invariant
signal-to-distortion ratio (SI-SDR) and 13.1 dB SDR, revealing new
possibilities for deep learning based phase reconstruction and representing a
fundamental progress towards solving the notoriously-hard cocktail party
problem.
|
Sparse Persistent RNNs: Squeezing Large Recurrent Networks On-Chip | Recurrent Neural Networks (RNNs) are powerful tools for solving
sequence-based problems, but their efficacy and execution time are dependent on
the size of the network. Following recent work in simplifying these networks
with model pruning and a novel mapping of work onto GPUs, we design an
efficient implementation for sparse RNNs. We investigate several optimizations
and tradeoffs: Lamport timestamps, wide memory loads, and a bank-aware weight
layout. With these optimizations, we achieve speedups of over 6x over the next
best algorithm for a hidden layer of size 2304, batch size of 4, and a density
of 30%. Further, our technique allows for models of over 5x the size to fit on
a GPU for a speedup of 2x, enabling larger networks to help advance the
state-of-the-art. We perform case studies on NMT and speech recognition tasks
in the appendix, accelerating their recurrent layers by up to 3x.
|
From Principal Subspaces to Principal Components with Linear
Autoencoders | The autoencoder is an effective unsupervised learning model which is widely
used in deep learning. It is well known that an autoencoder with a single
fully-connected hidden layer, a linear activation function and a squared error
cost function trains weights that span the same subspace as the one spanned by
the principal component loading vectors, but that they are not identical to the
loading vectors. In this paper, we show how to recover the loading vectors from
the autoencoder weights.
|
Tensor Methods for Nonlinear Matrix Completion | In the low-rank matrix completion (LRMC) problem, the low-rank assumption
means that the columns (or rows) of the matrix to be completed are points on a
low-dimensional linear algebraic variety. This paper extends this thinking to
cases where the columns are points on a low-dimensional nonlinear algebraic
variety, a problem we call Low Algebraic Dimension Matrix Completion (LADMC).
Matrices whose columns belong to a union of subspaces are an important special
case. We propose a LADMC algorithm that leverages existing LRMC methods on a
tensorized representation of the data. For example, a second-order tensorized
representation is formed by taking the Kronecker product of each column with
itself, and we consider higher order tensorizations as well. This approach will
succeed in many cases where traditional LRMC is guaranteed to fail because the
data are low-rank in the tensorized representation but not in the original
representation. We provide a formal mathematical justification for the success
of our method. In particular, we give bounds of the rank of these data in the
tensorized representation, and we prove sampling requirements to guarantee
uniqueness of the solution. We also provide experimental results showing that
the new approach outperforms existing state-of-the-art methods for matrix
completion under a union of subspaces model.
|
Network Transplanting | This paper focuses on a new task, i.e., transplanting a
category-and-task-specific neural network to a generic, modular network without
strong supervision. We design an functionally interpretable structure for the
generic network. Like building LEGO blocks, we teach the generic network a new
category by directly transplanting the module corresponding to the category
from a pre-trained network with a few or even without sample annotations. Our
method incrementally adds new categories to the generic network but does not
affect representations of existing categories. In this way, our method breaks
the typical bottleneck of learning a net for massive tasks and categories, i.e.
the requirement of collecting samples for all tasks and categories at the same
time before the learning begins. Thus, we use a new distillation algorithm,
namely back-distillation, to overcome specific challenges of network
transplanting. Our method without training samples even outperformed the
baseline with 100 training samples.
|
A telescoping Bregmanian proximal gradient method without the global
Lipschitz continuity assumption | The problem of minimization of the sum of two convex functions has various
theoretical and real-world applications. One of the popular methods for solving
this problem is the proximal gradient method (proximal forward-backward
algorithm). A very common assumption in the use of this method is that the
gradient of the smooth term is globally Lipschitz continuous. However, this
assumption is not always satisfied in practice, thus casting a limitation on
the method. In this paper, we discuss, in a wide class of finite and
infinite-dimensional spaces, a new variant of the proximal gradient method
which does not impose the above-mentioned global Lipschitz continuity
assumption. A key contribution of the method is the dependence of the iterative
steps on a certain telescopic decomposition of the constraint set into subsets.
Moreover, we use a Bregman divergence in the proximal forward-backward
operation. Under certain practical conditions, a non-asymptotic rate of
convergence (that is, in the function values) is established, as well as the
weak convergence of the whole sequence to a minimizer. We also obtain a few
auxiliary results of independent interest.
|
Adaptive Sensing for Learning Nonstationary Environment Models | Most environmental phenomena, such as wind profiles, ozone concentration and
sunlight distribution under a forest canopy, exhibit nonstationary dynamics
i.e. phenomenon variation change depending on the location and time of
occurrence. Non-stationary dynamics pose both theoretical and practical
challenges to statistical machine learning algorithms aiming to accurately
capture the complexities governing the evolution of such processes. In this
paper, we address the sampling aspects of the problem of learning nonstationary
spatio-temporal models, and propose an efficient yet simple algorithm - LISAL.
The core idea in LISAL is to learn two models using Gaussian processes (GPs)
wherein the first is a nonstationary GP directly modeling the phenomenon. The
second model uses a stationary GP representing a latent space corresponding to
changes in dynamics, or the nonstationarity characteristics of the first model.
LISAL involves adaptively sampling the latent space dynamics using information
theory quantities to reduce the computational cost during the learning phase.
The relevance of LISAL is extensively validated using multiple real world
datasets.
|
Distributed Differentially-Private Algorithms for Matrix and Tensor
Factorization | In many signal processing and machine learning applications, datasets
containing private information are held at different locations, requiring the
development of distributed privacy-preserving algorithms. Tensor and matrix
factorizations are key components of many processing pipelines. In the
distributed setting, differentially private algorithms suffer because they
introduce noise to guarantee privacy. This paper designs new and improved
distributed and differentially private algorithms for two popular matrix and
tensor factorization methods: principal component analysis (PCA) and orthogonal
tensor decomposition (OTD). The new algorithms employ a correlated noise design
scheme to alleviate the effects of noise and can achieve the same noise level
as the centralized scenario. Experiments on synthetic and real data illustrate
the regimes in which the correlated noise allows performance matching with the
centralized setting, outperforming previous methods and demonstrating that
meaningful utility is possible while guaranteeing differential privacy.
|
Efficiently Learning Nonstationary Gaussian Processes for Real World
Impact | Most real world phenomena such as sunlight distribution under a forest
canopy, minerals concentration, stock valuation, exhibit nonstationary dynamics
i.e. phenomenon variation changes depending on the locality. Nonstationary
dynamics pose both theoretical and practical challenges to statistical machine
learning algorithms that aim to accurately capture the complexities governing
the evolution of such processes. Typically the nonstationary dynamics are
modeled using nonstationary Gaussian Process models (NGPS) that employ local
latent dynamics parameterization to correspondingly model the nonstationary
real observable dynamics. Recently, an approach based on most likely induced
latent dynamics representation attracted research community's attention for a
while. The approach could not be employed for large scale real world
applications because learning a most likely latent dynamics representation
involves maximization of marginal likelihood of the observed real dynamics that
becomes intractable as the number of induced latent points grows with problem
size. We have established a direct relationship between informativeness of the
induced latent dynamics and the marginal likelihood of the observed real
dynamics. This opens up the possibility of maximizing marginal likelihood of
observed real dynamics indirectly by near optimally maximizing entropy or
mutual information gain on the induced latent dynamics using greedy algorithms.
Therefore, for an efficient yet accurate inference, we propose to build an
induced latent dynamics representation using a novel algorithm LISAL that
adaptively maximizes entropy or mutual information on the induced latent
dynamics and marginal likelihood of observed real dynamics in an iterative
manner. The relevance of LISAL is validated using real world datasets.
|
Classification of auditory stimuli from EEG signals with a regulated
recurrent neural network reservoir | The use of electroencephalogram (EEG) as the main input signal in
brain-machine interfaces has been widely proposed due to the non-invasive
nature of the EEG. Here we are specifically interested in interfaces that
extract information from the auditory system and more specifically in the task
of classifying heard speech from EEGs. To do so, we propose to limit the
preprocessing of the EEGs and use machine learning approaches to automatically
extract their meaningful characteristics. More specifically, we use a regulated
recurrent neural network (RNN) reservoir, which has been shown to outperform
classic machine learning approaches when applied to several different
bio-signals, and we compare it with a deep neural network approach. Moreover,
we also investigate the classification performance as a function of the number
of EEG electrodes. A set of 8 subjects were presented randomly with 3 different
auditory stimuli (English vowels a, i and u). We obtained an excellent
classification rate of 83.2% with the RNN when considering all 64 electrodes. A
rate of 81.7% was achieved with only 10 electrodes.
|
Scalable Bilinear $\pi$ Learning Using State and Action Features | Approximate linear programming (ALP) represents one of the major algorithmic
families to solve large-scale Markov decision processes (MDP). In this work, we
study a primal-dual formulation of the ALP, and develop a scalable, model-free
algorithm called bilinear $\pi$ learning for reinforcement learning when a
sampling oracle is provided. This algorithm enjoys a number of advantages.
First, it adopts (bi)linear models to represent the high-dimensional value
function and state-action distributions, using given state and action features.
Its run-time complexity depends on the number of features, not the size of the
underlying MDPs. Second, it operates in a fully online fashion without having
to store any sample, thus having minimal memory footprint. Third, we prove that
it is sample-efficient, solving for the optimal policy to high precision with a
sample complexity linear in the dimension of the parameter space.
|
Event Forecasting with Pattern Markov Chains | We present a system for online probabilistic event forecasting. We assume
that a user is interested in detecting and forecasting event patterns, given in
the form of regular expressions. Our system can consume streams of events and
forecast when the pattern is expected to be fully matched. As more events are
consumed, the system revises its forecasts to reflect possible changes in the
state of the pattern. The framework of Pattern Markov Chains is used in order
to learn a probabilistic model for the pattern, with which forecasts with
guaranteed precision may be produced, in the form of intervals within which a
full match is expected. Experimental results from real-world datasets are shown
and the quality of the produced forecasts is explored, using both precision
scores and two other metrics: spread, which refers to the "focusing resolution"
of a forecast (interval length), and distance, which captures how early a
forecast is reported.
|
Automatic classification of trees using a UAV onboard camera and deep
learning | Automatic classification of trees using remotely sensed data has been a dream
of many scientists and land use managers. Recently, Unmanned aerial vehicles
(UAV) has been expected to be an easy-to-use, cost-effective tool for remote
sensing of forests, and deep learning has attracted attention for its ability
concerning machine vision. In this study, using a commercially available UAV
and a publicly available package for deep learning, we constructed a machine
vision system for the automatic classification of trees. In our method, we
segmented a UAV photography image of forest into individual tree crowns and
carried out object-based deep learning. As a result, the system was able to
classify 7 tree types at 89.0% accuracy. This performance is notable because we
only used basic RGB images from a standard UAV. In contrast, most of previous
studies used expensive hardware such as multispectral imagers to improve the
performance. This result means that our method has the potential to classify
individual trees in a cost-effective manner. This can be a usable tool for many
forest researchers and managements.
|
Mining within-trial oscillatory brain dynamics to address the
variability of optimized spatial filters | Data-driven spatial filtering algorithms optimize scores such as the contrast
between two conditions to extract oscillatory brain signal components. Most
machine learning approaches for filter estimation, however, disregard
within-trial temporal dynamics and are extremely sensitive to changes in
training data and involved hyperparameters. This leads to highly variable
solutions and impedes the selection of a suitable candidate for,
e.g.,~neurotechnological applications. Fostering component introspection, we
propose to embrace this variability by condensing the functional signatures of
a large set of oscillatory components into homogeneous clusters, each
representing specific within-trial envelope dynamics.
The proposed method is exemplified by and evaluated on a complex hand force
task with a rich within-trial structure. Based on electroencephalography data
of 18 healthy subjects, we found that the components' distinct temporal
envelope dynamics are highly subject-specific. On average, we obtained seven
clusters per subject, which were strictly confined regarding their underlying
frequency bands. As the analysis method is not limited to a specific spatial
filtering algorithm, it could be utilized for a wide range of
neurotechnological applications, e.g., to select and monitor functionally
relevant features for brain-computer interface protocols in stroke
rehabilitation.
|
Offline Evaluation of Ranking Policies with Click Models | Many web systems rank and present a list of items to users, from recommender
systems to search and advertising. An important problem in practice is to
evaluate new ranking policies offline and optimize them before they are
deployed. We address this problem by proposing evaluation algorithms for
estimating the expected number of clicks on ranked lists from historical logged
data. The existing algorithms are not guaranteed to be statistically efficient
in our problem because the number of recommended lists can grow exponentially
with their length. To overcome this challenge, we use models of user
interaction with the list of items, the so-called click models, to construct
estimators that learn statistically efficiently. We analyze our estimators and
prove that they are more efficient than the estimators that do not use the
structure of the click model, under the assumption that the click model holds.
We evaluate our estimators in a series of experiments on a real-world dataset
and show that they consistently outperform prior estimators.
|
Deep Reinforcement Learning to Acquire Navigation Skills for
Wheel-Legged Robots in Complex Environments | Mobile robot navigation in complex and dynamic environments is a challenging
but important problem. Reinforcement learning approaches fail to solve these
tasks efficiently due to reward sparsities, temporal complexities and
high-dimensionality of sensorimotor spaces which are inherent in such problems.
We present a novel approach to train action policies to acquire navigation
skills for wheel-legged robots using deep reinforcement learning. The policy
maps height-map image observations to motor commands to navigate to a target
position while avoiding obstacles. We propose to acquire the multifaceted
navigation skill by learning and exploiting a number of manageable navigation
behaviors. We also introduce a domain randomization technique to improve the
versatility of the training samples. We demonstrate experimentally a
significant improvement in terms of data-efficiency, success rate, robustness
against irrelevant sensory data, and also the quality of the maneuver skills.
|
Using Machine Learning to Improve Cylindrical Algebraic Decomposition | Cylindrical Algebraic Decomposition (CAD) is a key tool in computational
algebraic geometry, best known as a procedure to enable Quantifier Elimination
over real-closed fields. However, it has a worst case complexity doubly
exponential in the size of the input, which is often encountered in practice.
It has been observed that for many problems a change in algorithm settings or
problem formulation can cause huge differences in runtime costs, changing
problem instances from intractable to easy. A number of heuristics have been
developed to help with such choices, but the complicated nature of the
geometric relationships involved means these are imperfect and can sometimes
make poor choices. We investigate the use of machine learning (specifically
support vector machines) to make such choices instead.
Machine learning is the process of fitting a computer model to a complex
function based on properties learned from measured data. In this paper we apply
it in two case studies: the first to select between heuristics for choosing a
CAD variable ordering; the second to identify when a CAD problem instance would
benefit from Groebner Basis preconditioning. These appear to be the first such
applications of machine learning to Symbolic Computation. We demonstrate in
both cases that the machine learned choice outperforms human developed
heuristics.
|
Learning Non-Stationary Space-Time Models for Environmental Monitoring | One of the primary aspects of sustainable development involves accurate
understanding and modeling of environmental phenomena. Many of these phenomena
exhibit variations in both space and time and it is imperative to develop a
deeper understanding of techniques that can model space-time dynamics
accurately. In this paper we propose NOSTILL-GP - NOn-stationary Space TIme
variable Latent Length scale GP, a generic non-stationary, spatio-temporal
Gaussian Process (GP) model. We present several strategies, for efficient
training of our model, necessary for real-world applicability. Extensive
empirical validation is performed using three real-world environmental
monitoring datasets, with diverse dynamics across space and time. Results from
the experiments clearly demonstrate general applicability and effectiveness of
our approach for applications in environmental monitoring.
|
Persistent Monitoring of Stochastic Spatio-temporal Phenomena with a
Small Team of Robots | This paper presents a solution for persistent monitoring of real-world
stochastic phenomena, where the underlying covariance structure changes sharply
across time, using a small number of mobile robot sensors. We propose an
adaptive solution for the problem where stochastic real-world dynamics are
modeled as a Gaussian Process (GP). The belief on the underlying covariance
structure is learned from recently observed dynamics as a Gaussian Mixture (GM)
in the low-dimensional hyper-parameters space of the GP and adapted across time
using Sequential Monte Carlo methods. Each robot samples a belief point from
the GM and locally optimizes a set of informative regions by greedy
maximization of the submodular entropy function. The key contributions of this
paper are threefold: adapting the belief on the covariance using Markov Chain
Monte Carlo (MCMC) sampling such that particles survive even under sharp
covariance changes across time; exploiting the belief to transform the problem
of entropy maximization into a decentralized one; and developing an
approximation algorithm to maximize entropy on a set of informative regions in
the continuous space. We illustrate the application of the proposed solution
through extensive simulations using an artificial dataset and multiple real
datasets from fixed sensor deployments, and compare it to three competing
state-of-the-art approaches.
|
Decoupled Parallel Backpropagation with Convergence Guarantee | Backpropagation algorithm is indispensable for the training of feedforward
neural networks. It requires propagating error gradients sequentially from the
output layer all the way back to the input layer. The backward locking in
backpropagation algorithm constrains us from updating network layers in
parallel and fully leveraging the computing resources. Recently, several
algorithms have been proposed for breaking the backward locking. However, their
performances degrade seriously when networks are deep. In this paper, we
propose decoupled parallel backpropagation algorithm for deep learning
optimization with convergence guarantee. Firstly, we decouple the
backpropagation algorithm using delayed gradients, and show that the backward
locking is removed when we split the networks into multiple modules. Then, we
utilize decoupled parallel backpropagation in two stochastic methods and prove
that our method guarantees convergence to critical points for the non-convex
problem. Finally, we perform experiments for training deep convolutional neural
networks on benchmark datasets. The experimental results not only confirm our
theoretical analysis, but also demonstrate that the proposed method can achieve
significant speedup without loss of accuracy.
|
An improvement of the convergence proof of the ADAM-Optimizer | A common way to train neural networks is the Backpropagation. This algorithm
includes a gradient descent method, which needs an adaptive step size. In the
area of neural networks, the ADAM-Optimizer is one of the most popular adaptive
step size methods. It was invented in \cite{Kingma.2015} by Kingma and Ba. The
$5865$ citations in only three years shows additionally the importance of the
given paper. We discovered that the given convergence proof of the optimizer
contains some mistakes, so that the proof will be wrong. In this paper we give
an improvement to the convergence proof of the ADAM-Optimizer.
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.