title
stringlengths 7
246
| abstract
stringlengths 6
3.31k
|
|---|---|
Towards Robust Neural Networks via Random Self-ensemble | Recent studies have revealed the vulnerability of deep neural networks: A
small adversarial perturbation that is imperceptible to human can easily make a
well-trained deep neural network misclassify. This makes it unsafe to apply
neural networks in security-critical applications. In this paper, we propose a
new defense algorithm called Random Self-Ensemble (RSE) by combining two
important concepts: {\bf randomness} and {\bf ensemble}. To protect a targeted
model, RSE adds random noise layers to the neural network to prevent the strong
gradient-based attacks, and ensembles the prediction over random noises to
stabilize the performance. We show that our algorithm is equivalent to ensemble
an infinite number of noisy models $f_\epsilon$ without any additional memory
overhead, and the proposed training procedure based on noisy stochastic
gradient descent can ensure the ensemble model has a good predictive
capability. Our algorithm significantly outperforms previous defense techniques
on real data sets. For instance, on CIFAR-10 with VGG network (which has 92\%
accuracy without any attack), under the strong C\&W attack within a certain
distortion tolerance, the accuracy of unprotected model drops to less than
10\%, the best previous defense technique has $48\%$ accuracy, while our method
still has $86\%$ prediction accuracy under the same level of attack. Finally,
our method is simple and easy to integrate into any neural network.
|
GANGs: Generative Adversarial Network Games | Generative Adversarial Networks (GAN) have become one of the most successful
frameworks for unsupervised generative modeling. As GANs are difficult to train
much research has focused on this. However, very little of this research has
directly exploited game-theoretic techniques. We introduce Generative
Adversarial Network Games (GANGs), which explicitly model a finite zero-sum
game between a generator ($G$) and classifier ($C$) that use mixed strategies.
The size of these games precludes exact solution methods, therefore we define
resource-bounded best responses (RBBRs), and a resource-bounded Nash
Equilibrium (RB-NE) as a pair of mixed strategies such that neither $G$ or $C$
can find a better RBBR. The RB-NE solution concept is richer than the notion of
`local Nash equilibria' in that it captures not only failures of escaping local
optima of gradient descent, but applies to any approximate best response
computations, including methods with random restarts. To validate our approach,
we solve GANGs with the Parallel Nash Memory algorithm, which provably
monotonically converges to an RB-NE. We compare our results to standard GAN
setups, and demonstrate that our method deals well with typical GAN problems
such as mode collapse, partial mode coverage and forgetting.
|
Improving Network Robustness against Adversarial Attacks with Compact
Convolution | Though Convolutional Neural Networks (CNNs) have surpassed human-level
performance on tasks such as object classification and face verification, they
can easily be fooled by adversarial attacks. These attacks add a small
perturbation to the input image that causes the network to misclassify the
sample. In this paper, we focus on neutralizing adversarial attacks by compact
feature learning. In particular, we show that learning features in a closed and
bounded space improves the robustness of the network. We explore the effect of
L2-Softmax Loss, that enforces compactness in the learned features, thus
resulting in enhanced robustness to adversarial perturbations. Additionally, we
propose compact convolution, a novel method of convolution that when
incorporated in conventional CNNs improves their robustness. Compact
convolution ensures feature compactness at every layer such that they are
bounded and close to each other. Extensive experiments show that Compact
Convolutional Networks (CCNs) neutralize multiple types of attacks, and perform
better than existing methods in defending adversarial attacks, without
incurring any additional training overhead compared to CNNs.
|
Spatial PixelCNN: Generating Images from Patches | In this paper we propose Spatial PixelCNN, a conditional autoregressive model
that generates images from small patches. By conditioning on a grid of pixel
coordinates and global features extracted from a Variational Autoencoder (VAE),
we are able to train on patches of images, and reproduce the full-sized image.
We show that it not only allows for generating high quality samples at the same
resolution as the underlying dataset, but is also capable of upscaling images
to arbitrary resolutions (tested at resolutions up to $50\times$) on the MNIST
dataset. Compared to a PixelCNN++ baseline, Spatial PixelCNN quantitatively and
qualitatively achieves similar performance on the MNIST dataset.
|
Sentiment Classification using Images and Label Embeddings | In this project we analysed how much semantic information images carry, and
how much value image data can add to sentiment analysis of the text associated
with the images. To better understand the contribution from images, we compared
models which only made use of image data, models which only made use of text
data, and models which combined both data types. We also analysed if this
approach could help sentiment classifiers generalize to unknown sentiments.
|
SHINE: Signed Heterogeneous Information Network Embedding for Sentiment
Link Prediction | In online social networks people often express attitudes towards others,
which forms massive sentiment links among users. Predicting the sign of
sentiment links is a fundamental task in many areas such as personal
advertising and public opinion analysis. Previous works mainly focus on textual
sentiment classification, however, text information can only disclose the "tip
of the iceberg" about users' true opinions, of which the most are unobserved
but implied by other sources of information such as social relation and users'
profile. To address this problem, in this paper we investigate how to predict
possibly existing sentiment links in the presence of heterogeneous information.
First, due to the lack of explicit sentiment links in mainstream social
networks, we establish a labeled heterogeneous sentiment dataset which consists
of users' sentiment relation, social relation and profile knowledge by
entity-level sentiment extraction method. Then we propose a novel and flexible
end-to-end Signed Heterogeneous Information Network Embedding (SHINE) framework
to extract users' latent representations from heterogeneous networks and
predict the sign of unobserved sentiment links. SHINE utilizes multiple deep
autoencoders to map each user into a low-dimension feature space while
preserving the network structure. We demonstrate the superiority of SHINE over
state-of-the-art baselines on link prediction and node recommendation in two
real-world datasets. The experimental results also prove the efficacy of SHINE
in cold start scenario.
|
Gradient Descent Learns One-hidden-layer CNN: Don't be Afraid of
Spurious Local Minima | We consider the problem of learning a one-hidden-layer neural network with
non-overlapping convolutional layer and ReLU activation, i.e., $f(\mathbf{Z},
\mathbf{w}, \mathbf{a}) = \sum_j a_j\sigma(\mathbf{w}^T\mathbf{Z}_j)$, in which
both the convolutional weights $\mathbf{w}$ and the output weights $\mathbf{a}$
are parameters to be learned. When the labels are the outputs from a teacher
network of the same architecture with fixed weights $(\mathbf{w}^*,
\mathbf{a}^*)$, we prove that with Gaussian input $\mathbf{Z}$, there is a
spurious local minimizer. Surprisingly, in the presence of the spurious local
minimizer, gradient descent with weight normalization from randomly initialized
weights can still be proven to recover the true parameters with constant
probability, which can be boosted to probability $1$ with multiple restarts. We
also show that with constant probability, the same procedure could also
converge to the spurious local minimum, showing that the local minimum plays a
non-trivial role in the dynamics of gradient descent. Furthermore, a
quantitative analysis shows that the gradient descent dynamics has two phases:
it starts off slow, but converges much faster after several iterations.
|
Exact upper and lower bounds on the misclassification probability | Exact lower and upper bounds on the best possible misclassification
probability for a finite number of classes are obtained in terms of the total
variation norms of the differences between the sub-distributions over the
classes. These bounds are compared with the exact bounds in terms of the
conditional entropy obtained by Feder and Merhav.
|
Tensor Train Neighborhood Preserving Embedding | In this paper, we propose a Tensor Train Neighborhood Preserving Embedding
(TTNPE) to embed multi-dimensional tensor data into low dimensional tensor
subspace. Novel approaches to solve the optimization problem in TTNPE are
proposed. For this embedding, we evaluate novel trade-off gain among
classification, computation, and dimensionality reduction (storage) for
supervised learning. It is shown that compared to the state-of-the-arts tensor
embedding methods, TTNPE achieves superior trade-off in classification,
computation, and dimensionality reduction in MNIST handwritten digits and
Weizmann face datasets.
|
Large-scale analysis of disease pathways in the human interactome | Discovering disease pathways, which can be defined as sets of proteins
associated with a given disease, is an important problem that has the potential
to provide clinically actionable insights for disease diagnosis, prognosis, and
treatment. Computational methods aid the discovery by relying on
protein-protein interaction (PPI) networks. They start with a few known
disease-associated proteins and aim to find the rest of the pathway by
exploring the PPI network around the known disease proteins. However, the
success of such methods has been limited, and failure cases have not been well
understood. Here we study the PPI network structure of 519 disease pathways. We
find that 90% of pathways do not correspond to single well-connected components
in the PPI network. Instead, proteins associated with a single disease tend to
form many separate connected components/regions in the network. We then
evaluate state-of-the-art disease pathway discovery methods and show that their
performance is especially poor on diseases with disconnected pathways. Thus, we
conclude that network connectivity structure alone may not be sufficient for
disease pathway discovery. However, we show that higher-order network
structures, such as small subgraphs of the pathway, provide a promising
direction for the development of new methods.
|
Raw Waveform-based Audio Classification Using Sample-level CNN
Architectures | Music, speech, and acoustic scene sound are often handled separately in the
audio domain because of their different signal characteristics. However, as the
image domain grows rapidly by versatile image classification models, it is
necessary to study extensible classification models in the audio domain as
well. In this study, we approach this problem using two types of sample-level
deep convolutional neural networks that take raw waveforms as input and uses
filters with small granularity. One is a basic model that consists of
convolution and pooling layers. The other is an improved model that
additionally has residual connections, squeeze-and-excitation modules and
multi-level concatenation. We show that the sample-level models reach
state-of-the-art performance levels for the three different categories of
sound. Also, we visualize the filters along layers and compare the
characteristics of learned filters.
|
Data Dropout in Arbitrary Basis for Deep Network Regularization | An important problem in training deep networks with high capacity is to
ensure that the trained network works well when presented with new inputs
outside the training dataset. Dropout is an effective regularization technique
to boost the network generalization in which a random subset of the elements of
the given data and the extracted features are set to zero during the training
process. In this paper, a new randomized regularization technique in which we
withhold a random part of the data without necessarily turning off the
neurons/data-elements is proposed. In the proposed method, of which the
conventional dropout is shown to be a special case, random data dropout is
performed in an arbitrary basis, hence the designation Generalized Dropout. We
also present a framework whereby the proposed technique can be applied
efficiently to convolutional neural networks. The presented numerical
experiments demonstrate that the proposed technique yields notable performance
gain. Generalized Dropout provides new insight into the idea of dropout, shows
that we can achieve different performance gains by using different bases
matrices, and opens up a new research question as of how to choose optimal
bases matrices that achieve maximal performance gain.
|
Deep Learning Diffuse Optical Tomography | Diffuse optical tomography (DOT) has been investigated as an alternative
imaging modality for breast cancer detection thanks to its excellent contrast
to hemoglobin oxidization level. However, due to the complicated non-linear
photon scattering physics and ill-posedness, the conventional reconstruction
algorithms are sensitive to imaging parameters such as boundary conditions. To
address this, here we propose a novel deep learning approach that learns
non-linear photon scattering physics and obtains an accurate three dimensional
(3D) distribution of optical anomalies. In contrast to the traditional
black-box deep learning approaches, our deep network is designed to invert the
Lippman-Schwinger integral equation using the recent mathematical theory of
deep convolutional framelets. As an example of clinical relevance, we applied
the method to our prototype DOT system. We show that our deep neural network,
trained with only simulation data, can accurately recover the location of
anomalies within biomimetic phantoms and live animals without the use of an
exogenous contrast agent.
|
Learning Multi-Level Hierarchies with Hindsight | Hierarchical agents have the potential to solve sequential decision making
tasks with greater sample efficiency than their non-hierarchical counterparts
because hierarchical agents can break down tasks into sets of subtasks that
only require short sequences of decisions. In order to realize this potential
of faster learning, hierarchical agents need to be able to learn their multiple
levels of policies in parallel so these simpler subproblems can be solved
simultaneously. Yet, learning multiple levels of policies in parallel is hard
because it is inherently unstable: changes in a policy at one level of the
hierarchy may cause changes in the transition and reward functions at higher
levels in the hierarchy, making it difficult to jointly learn multiple levels
of policies. In this paper, we introduce a new Hierarchical Reinforcement
Learning (HRL) framework, Hierarchical Actor-Critic (HAC), that can overcome
the instability issues that arise when agents try to jointly learn multiple
levels of policies. The main idea behind HAC is to train each level of the
hierarchy independently of the lower levels by training each level as if the
lower level policies are already optimal. We demonstrate experimentally in both
grid world and simulated robotics domains that our approach can significantly
accelerate learning relative to other non-hierarchical and hierarchical
methods. Indeed, our framework is the first to successfully learn 3-level
hierarchies in parallel in tasks with continuous state and action spaces.
|
Learning Independent Causal Mechanisms | Statistical learning relies upon data sampled from a distribution, and we
usually do not care what actually generated it in the first place. From the
point of view of causal modeling, the structure of each distribution is induced
by physical mechanisms that give rise to dependences between observables.
Mechanisms, however, can be meaningful autonomous modules of generative models
that make sense beyond a particular entailed data distribution, lending
themselves to transfer between problems. We develop an algorithm to recover a
set of independent (inverse) mechanisms from a set of transformed data points.
The approach is unsupervised and based on a set of experts that compete for
data generated by the mechanisms, driving specialization. We analyze the
proposed method in a series of experiments on image data. Each expert learns to
map a subset of the transformed data back to a reference distribution. The
learned mechanisms generalize to novel domains. We discuss implications for
transfer learning and links to recent trends in generative modeling.
|
Temporal Attention augmented Bilinear Network for Financial Time-Series
Data Analysis | Financial time-series forecasting has long been a challenging problem because
of the inherently noisy and stochastic nature of the market. In the
High-Frequency Trading (HFT), forecasting for trading purposes is even a more
challenging task since an automated inference system is required to be both
accurate and fast. In this paper, we propose a neural network layer
architecture that incorporates the idea of bilinear projection as well as an
attention mechanism that enables the layer to detect and focus on crucial
temporal information. The resulting network is highly interpretable, given its
ability to highlight the importance and contribution of each temporal instance,
thus allowing further analysis on the time instances of interest. Our
experiments in a large-scale Limit Order Book (LOB) dataset show that a
two-hidden-layer network utilizing our proposed layer outperforms by a large
margin all existing state-of-the-art results coming from much deeper
architectures while requiring far fewer computations.
|
Chord Generation from Symbolic Melody Using BLSTM Networks | Generating a chord progression from a monophonic melody is a challenging
problem because a chord progression requires a series of layered notes played
simultaneously. This paper presents a novel method of generating chord
sequences from a symbolic melody using bidirectional long short-term memory
(BLSTM) networks trained on a lead sheet database. To this end, a group of
feature vectors composed of 12 semitones is extracted from the notes in each
bar of monophonic melodies. In order to ensure that the data shares uniform key
and duration characteristics, the key and the time signatures of the vectors
are normalized. The BLSTM networks then learn from the data to incorporate the
temporal dependencies to produce a chord progression. Both quantitative and
qualitative evaluations are conducted by comparing the proposed method with the
conventional HMM and DNN-HMM based approaches. Proposed model achieves 23.8%
and 11.4% performance increase from the other models, respectively. User
studies further confirm that the chord sequences generated by the proposed
method are preferred by listeners.
|
Vprop: Variational Inference using RMSprop | Many computationally-efficient methods for Bayesian deep learning rely on
continuous optimization algorithms, but the implementation of these methods
requires significant changes to existing code-bases. In this paper, we propose
Vprop, a method for Gaussian variational inference that can be implemented with
two minor changes to the off-the-shelf RMSprop optimizer. Vprop also reduces
the memory requirements of Black-Box Variational Inference by half. We derive
Vprop using the conjugate-computation variational inference method, and
establish its connections to Newton's method, natural-gradient methods, and
extended Kalman filters. Overall, this paper presents Vprop as a principled,
computationally-efficient, and easy-to-implement method for Bayesian deep
learning.
|
Adaptive Quantization for Deep Neural Network | In recent years Deep Neural Networks (DNNs) have been rapidly developed in
various applications, together with increasingly complex architectures. The
performance gain of these DNNs generally comes with high computational costs
and large memory consumption, which may not be affordable for mobile platforms.
Deep model quantization can be used for reducing the computation and memory
costs of DNNs, and deploying complex DNNs on mobile equipment. In this work, we
propose an optimization framework for deep model quantization. First, we
propose a measurement to estimate the effect of parameter quantization errors
in individual layers on the overall model prediction accuracy. Then, we propose
an optimization process based on this measurement for finding optimal
quantization bit-width for each layer. This is the first work that
theoretically analyse the relationship between parameter quantization errors of
individual layers and model accuracy. Our new quantization algorithm
outperforms previous quantization optimization methods, and achieves 20-40%
higher compression rate compared to equal bit-width quantization at the same
model prediction accuracy.
|
Structured Deep Neural Network Pruning via Matrix Pivoting | Deep Neural Networks (DNNs) are the key to the state-of-the-art machine
vision, sensor fusion and audio/video signal processing. Unfortunately, their
computation complexity and tight resource constraints on the Edge make them
hard to leverage on mobile, embedded and IoT devices. Due to great diversity of
Edge devices, DNN designers have to take into account the hardware platform and
application requirements during network training. In this work we introduce
pruning via matrix pivoting as a way to improve network pruning by compromising
between the design flexibility of architecture-oblivious and performance
efficiency of architecture-aware pruning, the two dominant techniques for
obtaining resource-efficient DNNs. We also describe local and global network
optimization techniques for efficient implementation of the resulting pruned
networks. In combination, the proposed pruning and implementation result in
close to linear speed up with the reduction of network coefficients during
pruning.
|
Transferring Autonomous Driving Knowledge on Simulated and Real
Intersections | We view intersection handling on autonomous vehicles as a reinforcement
learning problem, and study its behavior in a transfer learning setting. We
show that a network trained on one type of intersection generally is not able
to generalize to other intersections. However, a network that is pre-trained on
one intersection and fine-tuned on another performs better on the new task
compared to training in isolation. This network also retains knowledge of the
prior task, even though some forgetting occurs. Finally, we show that the
benefits of fine-tuning hold when transferring simulated intersection handling
knowledge to a real autonomous vehicle.
|
Stochastic Maximum Likelihood Optimization via Hypernetworks | This work explores maximum likelihood optimization of neural networks through
hypernetworks. A hypernetwork initializes the weights of another network, which
in turn can be employed for typical functional tasks such as regression and
classification. We optimize hypernetworks to directly maximize the conditional
likelihood of target variables given input. Using this approach we obtain
competitive empirical results on regression and classification benchmarks.
|
Learning Fast and Slow: PROPEDEUTICA for Real-time Malware Detection | Existing malware detectors on safety-critical devices have difficulties in
runtime detection due to the performance overhead. In this paper, we introduce
PROPEDEUTICA, a framework for efficient and effective real-time malware
detection, leveraging the best of conventional machine learning (ML) and deep
learning (DL) techniques. In PROPEDEUTICA, all software start execution are
considered as benign and monitored by a conventional ML classifier for fast
detection. If the software receives a borderline classification from the ML
detector (e.g. the software is 50% likely to be benign and 50% likely to be
malicious), the software will be transferred to a more accurate, yet
performance demanding DL detector. To address spatial-temporal dynamics and
software execution heterogeneity, we introduce a novel DL architecture
(DEEPMALWARE) for PROPEDEUTICA with multi-stream inputs. We evaluated
PROPEDEUTICA with 9,115 malware samples and 1,338 benign software from various
categories for the Windows OS. With a borderline interval of [30%-70%],
PROPEDEUTICA achieves an accuracy of 94.34% and a false-positive rate of 8.75%,
with 41.45% of the samples moved for DEEPMALWARE analysis. Even using only CPU,
PROPEDEUTICA can detect malware within less than 0.1 seconds.
|
Statistical Inference for Incomplete Ranking Data: The Case of
Rank-Dependent Coarsening | We consider the problem of statistical inference for ranking data,
specifically rank aggregation, under the assumption that samples are incomplete
in the sense of not comprising all choice alternatives. In contrast to most
existing methods, we explicitly model the process of turning a full ranking
into an incomplete one, which we call the coarsening process. To this end, we
propose the concept of rank-dependent coarsening, which assumes that incomplete
rankings are produced by projecting a full ranking to a random subset of ranks.
For a concrete instantiation of our model, in which full rankings are drawn
from a Plackett-Luce distribution and observations take the form of pairwise
preferences, we study the performance of various rank aggregation methods. In
addition to predictive accuracy in the finite sample setting, we address the
theoretical question of consistency, by which we mean the ability to recover a
target ranking when the sample size goes to infinity, despite a potential bias
in the observations caused by the (unknown) coarsening.
|
Episodic memory for continual model learning | Both the human brain and artificial learning agents operating in real-world
or comparably complex environments are faced with the challenge of online model
selection. In principle this challenge can be overcome: hierarchical Bayesian
inference provides a principled method for model selection and it converges on
the same posterior for both off-line (i.e. batch) and online learning. However,
maintaining a parameter posterior for each model in parallel has in general an
even higher memory cost than storing the entire data set and is consequently
clearly unfeasible. Alternatively, maintaining only a limited set of models in
memory could limit memory requirements. However, sufficient statistics for one
model will usually be insufficient for fitting a different kind of model,
meaning that the agent loses information with each model change. We propose
that episodic memory can circumvent the challenge of limited memory-capacity
online model selection by retaining a selected subset of data points. We design
a method to compute the quantities necessary for model selection even when the
data is discarded and only statistics of one (or few) learnt models are
available. We demonstrate on a simple model that a limited-sized episodic
memory buffer, when the content is optimised to retain data with statistics not
matching the current representation, can resolve the fundamental challenge of
online model selection.
|
A dual framework for low-rank tensor completion | One of the popular approaches for low-rank tensor completion is to use the
latent trace norm regularization. However, most existing works in this
direction learn a sparse combination of tensors. In this work, we fill this gap
by proposing a variant of the latent trace norm that helps in learning a
non-sparse combination of tensors. We develop a dual framework for solving the
low-rank tensor completion problem. We first show a novel characterization of
the dual solution space with an interesting factorization of the optimal
solution. Overall, the optimal solution is shown to lie on a Cartesian product
of Riemannian manifolds. Furthermore, we exploit the versatile Riemannian
optimization framework for proposing computationally efficient trust region
algorithm. The experiments illustrate the efficacy of the proposed algorithm on
several real-world datasets across applications.
|
tHoops: A Multi-Aspect Analytical Framework Spatio-Temporal Basketball
Data | During the past few years advancements in sports information systems and
technology has allowed us to collect a number of detailed spatio-temporal data
capturing various aspects of basketball. For example, shot charts, that is,
maps capturing locations of (made or missed) shots, and spatio-temporal
trajectories for all the players on the court can capture information about the
offensive and defensive tendencies and schemes of a team. Characterization of
these processes is important for player and team comparisons, pre-game
scouting, game preparation etc. Playing tendencies among teams have
traditionally been compared in a heuristic manner. Recently automated ways for
similar comparisons have appeared in the sports analytics literature. However,
these approaches are almost exclusively focused on the spatial distribution of
the underlying actions (usually shots taken), ignoring a multitude of other
parameters that can affect the action studied. In this work, we propose a
framework based on tensor decomposition for obtaining a set of prototype
spatio-temporal patterns based on the core spatiotemporal information and
contextual meta-data. The core of our framework is a 3D tensor X, whose
dimensions represent the entity under consideration (team, player, possession
etc.), the location on the court and time. We make use of the PARAFAC
decomposition and we decompose the tensor into several interpretable patterns,
that can be thought of as prototype patterns of the process examined (e.g.,
shot selection, offensive schemes etc.). We also introduce an approach for
choosing the number of components to be considered. Using the tensor
components, we can then express every entity as a weighted combination of these
components. The framework introduced in this paper can have further
applications in the work-flow of the basketball operations of a franchise,
which we also briefly discuss.
|
On the Real-time Vehicle Placement Problem | Motivated by ride-sharing platforms' efforts to reduce their riders' wait
times for a vehicle, this paper introduces a novel problem of placing vehicles
to fulfill real-time pickup requests in a spatially and temporally changing
environment. The real-time nature of this problem makes it fundamentally
different from other placement and scheduling problems, as it requires not only
real-time placement decisions but also handling real-time request dynamics,
which are influenced by human mobility patterns. We use a dataset of ten
million ride requests from four major U.S. cities to show that the requests
exhibit significant self-similarity. We then propose distributed online
learning algorithms for the real-time vehicle placement problem and bound their
expected performance under this observed self-similarity.
|
Learning by Asking Questions | We introduce an interactive learning framework for the development and
testing of intelligent visual systems, called learning-by-asking (LBA). We
explore LBA in context of the Visual Question Answering (VQA) task. LBA differs
from standard VQA training in that most questions are not observed during
training time, and the learner must ask questions it wants answers to. Thus,
LBA more closely mimics natural learning and has the potential to be more
data-efficient than the traditional VQA setting. We present a model that
performs LBA on the CLEVR dataset, and show that it automatically discovers an
easy-to-hard curriculum when learning interactively from an oracle. Our LBA
generated data consistently matches or outperforms the CLEVR train data and is
more sample efficient. We also show that our model asks questions that
generalize to state-of-the-art VQA models and to novel test time distributions.
|
Clustering Stable Instances of Euclidean k-means | The Euclidean k-means problem is arguably the most widely-studied clustering
problem in machine learning. While the k-means objective is NP-hard in the
worst-case, practitioners have enjoyed remarkable success in applying
heuristics like Lloyd's algorithm for this problem. To address this disconnect,
we study the following question: what properties of real-world instances will
enable us to design efficient algorithms and prove guarantees for finding the
optimal clustering? We consider a natural notion called additive perturbation
stability that we believe captures many practical instances. Stable instances
have unique optimal k-means solutions that do not change even when each point
is perturbed a little (in Euclidean distance). This captures the property that
the k-means optimal solution should be tolerant to measurement errors and
uncertainty in the points. We design efficient algorithms that provably recover
the optimal clustering for instances that are additive perturbation stable.
When the instance has some additional separation, we show an efficient
algorithm with provable guarantees that is also robust to outliers. We
complement these results by studying the amount of stability in real datasets
and demonstrating that our algorithm performs well on these benchmark datasets.
|
An Equivalence of Fully Connected Layer and Convolutional Layer | This article demonstrates that convolutional operation can be converted to
matrix multiplication, which has the same calculation way with fully connected
layer. The article is helpful for the beginners of the neural network to
understand how fully connected layer and the convolutional layer work in the
backend. To be concise and to make the article more readable, we only consider
the linear case. It can be extended to the non-linear case easily through
plugging in a non-linear encapsulation to the values like this $\sigma(x)$
denoted as $x^{\prime}$.
|
Development of Statewide AADT Estimation Model from Short-Term Counts: A
Comparative Study for South Carolina | Annual Average Daily Traffic (AADT) is an important parameter used in traffic
engineering analysis. Departments of Transportation (DOTs) continually collect
traffic count using both permanent count stations (i.e., Automatic Traffic
Recorders or ATRs) and temporary short-term count stations. In South Carolina,
87% of the ATRs are located on interstates and arterial highways. For most
secondary highways (i.e., collectors and local roads), AADT is estimated based
on short-term counts. This paper develops AADT estimation models for different
roadway functional classes with two machine learning techniques: Artificial
Neural Network (ANN) and Support Vector Regression (SVR). The models aim to
predict AADT from short-term counts. The results are first compared against
each other to identify the best model. Then, the results of the best model are
compared against a regression method and factor-based method. The comparison
reveals the superiority of SVR for AADT estimation for different roadway
functional classes over all other methods. Among all developed models for
different functional roadway classes, the SVR-based model shows a minimum root
mean square error (RMSE) of 0.22 and a mean absolute percentage error (MAPE) of
11.3% for the interstate/expressway functional class. This model also shows a
higher R-squared value compared to the traditional factor-based model and
regression model. SVR models are validated for each roadway functional class
using the 2016 ATR data and selected short-term count data collected by the
South Carolina Department of Transportation (SCDOT). The validation results
show that the SVR-based AADT estimation models can be used by the SCDOT as a
reliable option to predict AADT from the short-term counts.
|
Compatibility Family Learning for Item Recommendation and Generation | Compatibility between items, such as clothes and shoes, is a major factor
among customer's purchasing decisions. However, learning "compatibility" is
challenging due to (1) broader notions of compatibility than those of
similarity, (2) the asymmetric nature of compatibility, and (3) only a small
set of compatible and incompatible items are observed. We propose an end-to-end
trainable system to embed each item into a latent vector and project a query
item into K compatible prototypes in the same space. These prototypes reflect
the broad notions of compatibility. We refer to both the embedding and
prototypes as "Compatibility Family". In our learned space, we introduce a
novel Projected Compatibility Distance (PCD) function which is differentiable
and ensures diversity by aiming for at least one prototype to be close to a
compatible item, whereas none of the prototypes are close to an incompatible
item. We evaluate our system on a toy dataset, two Amazon product datasets, and
Polyvore outfit dataset. Our method consistently achieves state-of-the-art
performance. Finally, we show that we can visualize the candidate compatible
prototypes using a Metric-regularized Conditional Generative Adversarial
Network (MrCGAN), where the input is a projected prototype and the output is a
generated image of a compatible item. We ask human evaluators to judge the
relative compatibility between our generated images and images generated by
CGANs conditioned directly on query items. Our generated images are
significantly preferred, with roughly twice the number of votes as others.
|
Layer-wise Learning of Stochastic Neural Networks with Information
Bottleneck | Information Bottleneck (IB) is a generalization of rate-distortion theory
that naturally incorporates compression and relevance trade-offs for learning.
Though the original IB has been extensively studied, there has not been much
understanding of multiple bottlenecks which better fit in the context of neural
networks. In this work, we propose Information Multi-Bottlenecks (IMBs) as an
extension of IB to multiple bottlenecks which has a direct application to
training neural networks by considering layers as multiple bottlenecks and
weights as parameterized encoders and decoders. We show that the multiple
optimality of IMB is not simultaneously achievable for stochastic encoders. We
thus propose a simple compromised scheme of IMB which in turn generalizes
maximum likelihood estimate (MLE) principle in the context of stochastic neural
networks. We demonstrate the effectiveness of IMB on classification tasks and
adversarial robustness in MNIST and CIFAR10.
|
A Deeper Look at Experience Replay | Recently experience replay is widely used in various deep reinforcement
learning (RL) algorithms, in this paper we rethink the utility of experience
replay. It introduces a new hyper-parameter, the memory buffer size, which
needs carefully tuning. However unfortunately the importance of this new
hyper-parameter has been underestimated in the community for a long time. In
this paper we did a systematic empirical study of experience replay under
various function representations. We showcase that a large replay buffer can
significantly hurt the performance. Moreover, we propose a simple O(1) method
to remedy the negative influence of a large replay buffer. We showcase its
utility in both simple grid world and challenging domains like Atari games.
|
Learning Sparse Neural Networks through $L_0$ Regularization | We propose a practical method for $L_0$ norm regularization for neural
networks: pruning the network during training by encouraging weights to become
exactly zero. Such regularization is interesting since (1) it can greatly speed
up training and inference, and (2) it can improve generalization. AIC and BIC,
well-known model selection criteria, are special cases of $L_0$ regularization.
However, since the $L_0$ norm of weights is non-differentiable, we cannot
incorporate it directly as a regularization term in the objective function. We
propose a solution through the inclusion of a collection of non-negative
stochastic gates, which collectively determine which weights to set to zero. We
show that, somewhat surprisingly, for certain distributions over the gates, the
expected $L_0$ norm of the resulting gated weights is differentiable with
respect to the distribution parameters. We further propose the \emph{hard
concrete} distribution for the gates, which is obtained by "stretching" a
binary concrete distribution and then transforming its samples with a
hard-sigmoid. The parameters of the distribution over the gates can then be
jointly optimized with the original network parameters. As a result our method
allows for straightforward and efficient learning of model structures with
stochastic gradient descent and allows for conditional computation in a
principled way. We perform various experiments to demonstrate the effectiveness
of the resulting approach and regularizer.
|
Examining Cooperation in Visual Dialog Models | In this work we propose a blackbox intervention method for visual dialog
models, with the aim of assessing the contribution of individual linguistic or
visual components. Concretely, we conduct structured or randomized
interventions that aim to impair an individual component of the model, and
observe changes in task performance. We reproduce a state-of-the-art visual
dialog model and demonstrate that our methodology yields surprising insights,
namely that both dialog and image information have minimal contributions to
task performance. The intervention method presented here can be applied as a
sanity check for the strength and robustness of each component in visual dialog
systems.
|
Linearly-Recurrent Autoencoder Networks for Learning Dynamics | This paper describes a method for learning low-dimensional approximations of
nonlinear dynamical systems, based on neural-network approximations of the
underlying Koopman operator. Extended Dynamic Mode Decomposition (EDMD)
provides a useful data-driven approximation of the Koopman operator for
analyzing dynamical systems. This paper addresses a fundamental problem
associated with EDMD: a trade-off between representational capacity of the
dictionary and over-fitting due to insufficient data. A new neural network
architecture combining an autoencoder with linear recurrent dynamics in the
encoded state is used to learn a low-dimensional and highly informative
Koopman-invariant subspace of observables. A method is also presented for
balanced model reduction of over-specified EDMD systems in feature space.
Nonlinear reconstruction using partially linear multi-kernel regression aims to
improve reconstruction accuracy from the low-dimensional state when the data
has complex but intrinsically low-dimensional structure. The techniques
demonstrate the ability to identify Koopman eigenfunctions of the unforced
Duffing equation, create accurate low-dimensional models of an unstable
cylinder wake flow, and make short-time predictions of the chaotic
Kuramoto-Sivashinsky equation.
|
Gaussian Process bandits with adaptive discretization | In this paper, the problem of maximizing a black-box function $f:\mathcal{X}
\to \mathbb{R}$ is studied in the Bayesian framework with a Gaussian Process
(GP) prior. In particular, a new algorithm for this problem is proposed, and
high probability bounds on its simple and cumulative regret are established.
The query point selection rule in most existing methods involves an exhaustive
search over an increasingly fine sequence of uniform discretizations of
$\mathcal{X}$. The proposed algorithm, in contrast, adaptively refines
$\mathcal{X}$ which leads to a lower computational complexity, particularly
when $\mathcal{X}$ is a subset of a high dimensional Euclidean space. In
addition to the computational gains, sufficient conditions are identified under
which the regret bounds of the new algorithm improve upon the known results.
Finally an extension of the algorithm to the case of contextual bandits is
proposed, and high probability bounds on the contextual regret are presented.
|
Learning to Fuse Music Genres with Generative Adversarial Dual Learning | FusionGAN is a novel genre fusion framework for music generation that
integrates the strengths of generative adversarial networks and dual learning.
In particular, the proposed method offers a dual learning extension that can
effectively integrate the styles of the given domains. To efficiently quantify
the difference among diverse domains and avoid the vanishing gradient issue,
FusionGAN provides a Wasserstein based metric to approximate the distance
between the target domain and the existing domains. Adopting the Wasserstein
distance, a new domain is created by combining the patterns of the existing
domains using adversarial learning. Experimental results on public music
datasets demonstrated that our approach could effectively merge two genres.
|
Deep linear neural networks with arbitrary loss: All local minima are
global | We consider deep linear networks with arbitrary convex differentiable loss.
We provide a short and elementary proof of the fact that all local minima are
global minima if the hidden layers are either 1) at least as wide as the input
layer, or 2) at least as wide as the output layer. This result is the strongest
possible in the following sense: If the loss is convex and Lipschitz but not
differentiable then deep linear networks can have sub-optimal local minima.
|
Learning Pain from Action Unit Combinations: A Weakly Supervised
Approach via Multiple Instance Learning | Patient pain can be detected highly reliably from facial expressions using a
set of facial muscle-based action units (AUs) defined by the Facial Action
Coding System (FACS). A key characteristic of facial expression of pain is the
simultaneous occurrence of pain-related AU combinations, whose automated
detection would be highly beneficial for efficient and practical pain
monitoring. Existing general Automated Facial Expression Recognition (AFER)
systems prove inadequate when applied specifically for detecting pain as they
either focus on detecting individual pain-related AUs but not on combinations
or they seek to bypass AU detection by training a binary pain classifier
directly on pain intensity data but are limited by lack of enough labeled data
for satisfactory training. In this paper, we propose a new approach that mimics
the strategy of human coders of decoupling pain detection into two consecutive
tasks: one performed at the individual video-frame level and the other at
video-sequence level. Using state-of-the-art AFER tools to detect single AUs at
the frame level, we propose two novel data structures to encode AU combinations
from single AU scores. Two weakly supervised learning frameworks namely
multiple instance learning (MIL) and multiple clustered instance learning
(MCIL) are employed corresponding to each data structure to learn pain from
video sequences. Experimental results show an 87% pain recognition accuracy
with 0.94 AUC (Area Under Curve) on the UNBC-McMaster Shoulder Pain Expression
dataset. Tests on long videos in a lung cancer patient video dataset
demonstrates the potential value of the proposed system for pain monitoring in
clinical settings.
|
Eigendecompositions of Transfer Operators in Reproducing Kernel Hilbert
Spaces | Transfer operators such as the Perron--Frobenius or Koopman operator play an
important role in the global analysis of complex dynamical systems. The
eigenfunctions of these operators can be used to detect metastable sets, to
project the dynamics onto the dominant slow processes, or to separate
superimposed signals. We extend transfer operator theory to reproducing kernel
Hilbert spaces and show that these operators are related to Hilbert space
representations of conditional distributions, known as conditional mean
embeddings in the machine learning community. Moreover, numerical methods to
compute empirical estimates of these embeddings are akin to data-driven methods
for the approximation of transfer operators such as extended dynamic mode
decomposition and its variants. One main benefit of the presented kernel-based
approaches is that these methods can be applied to any domain where a
similarity measure given by a kernel is available. We illustrate the results
with the aid of guiding examples and highlight potential applications in
molecular dynamics as well as video and text data analysis.
|
Deep learning for semantic segmentation of remote sensing images with
rich spectral content | With the rapid development of Remote Sensing acquisition techniques, there is
a need to scale and improve processing tools to cope with the observed increase
of both data volume and richness. Among popular techniques in remote sensing,
Deep Learning gains increasing interest but depends on the quality of the
training data. Therefore, this paper presents recent Deep Learning approaches
for fine or coarse land cover semantic segmentation estimation. Various 2D
architectures are tested and a new 3D model is introduced in order to jointly
process the spatial and spectral dimensions of the data. Such a set of networks
enables the comparison of the different spectral fusion schemes. Besides, we
also assess the use of a " noisy ground truth " (i.e. outdated and low spatial
resolution labels) for training and testing the networks.
|
Autonomous development and learning in artificial intelligence and
robotics: Scaling up deep learning to human--like learning | Autonomous lifelong development and learning is a fundamental capability of
humans, differentiating them from current deep learning systems. However, other
branches of artificial intelligence have designed crucial ingredients towards
autonomous learning: curiosity and intrinsic motivation, social learning and
natural interaction with peers, and embodiment. These mechanisms guide
exploration and autonomous choice of goals, and integrating them with deep
learning opens stimulating perspectives. Deep learning (DL) approaches made
great advances in artificial intelligence, but are still far away from human
learning. As argued convincingly by Lake et al., differences include human
capabilities to learn causal models of the world from very little data,
leveraging compositional representations and priors like intuitive physics and
psychology. However, there are other fundamental differences between current DL
systems and human learning, as well as technical ingredients to fill this gap,
that are either superficially, or not adequately, discussed by Lake et al.
These fundamental mechanisms relate to autonomous development and learning.
They are bound to play a central role in artificial intelligence in the future.
Current DL systems require engineers to manually specify a task-specific
objective function for every new task, and learn through off-line processing of
large training databases. On the contrary, humans learn autonomously open-ended
repertoires of skills, deciding for themselves which goals to pursue or value,
and which skills to explore, driven by intrinsic motivation/curiosity and
social learning through natural interaction with peers. Such learning processes
are incremental, online, and progressive. Human child development involves a
progressive increase of complexity in a curriculum of learning where skills are
explored, acquired, and built on each other, through particular ordering and
timing. Finally, human learning happens in the physical world, and through
bodily and physical experimentation, under severe constraints on energy, time,
and computational resources. In the two last decades, the field of
Developmental and Cognitive Robotics (Cangelosi and Schlesinger, 2015, Asada et
al., 2009), in strong interaction with developmental psychology and
neuroscience, has achieved significant advances in computational
|
On Deterministic Sampling Patterns for Robust Low-Rank Matrix Completion | In this letter, we study the deterministic sampling patterns for the
completion of low rank matrix, when corrupted with a sparse noise, also known
as robust matrix completion. We extend the recent results on the deterministic
sampling patterns in the absence of noise based on the geometric analysis on
the Grassmannian manifold. A special case where each column has a certain
number of noisy entries is considered, where our probabilistic analysis
performs very efficiently. Furthermore, assuming that the rank of the original
matrix is not given, we provide an analysis to determine if the rank of a valid
completion is indeed the actual rank of the data corrupted with sparse noise by
verifying some conditions.
|
Fully Convolutional Measurement Network for Compressive Sensing Image
Reconstruction | Recently, deep learning methods have made a significant improvement in
compressive sensing image reconstruction task. In the existing methods, the
scene is measured block by block due to the high computational complexity. This
results in block-effect of the recovered images. In this paper, we propose a
fully convolutional measurement network, where the scene is measured as a
whole. The proposed method powerfully removes the block-effect since the
structure information of scene images is preserved. To make the measure more
flexible, the measurement and the recovery parts are jointly trained. From the
experiments, it is shown that the results by the proposed method outperforms
those by the existing methods in PSNR, SSIM, and visual effect.
|
Three-Stream Convolutional Networks for Video-based Person
Re-Identification | This paper aims to develop a new architecture that can make full use of the
feature maps of convolutional networks. To this end, we study a number of
methods for video-based person re-identification and make the following
findings: 1) Max-pooling only focuses on the maximum value of a receptive
field, wasting a lot of information. 2) Networks with different streams even
including the one with the worst performance work better than networks with
same streams, where each one has the best performance alone. 3) A full
connection layer at the end of convolutional networks is not necessary. Based
on these studies, we propose a new convolutional architecture termed
Three-Stream Convolutional Networks (TSCN). It first uses different streams to
learn different aspects of feature maps for attentive spatio-temporal fusion of
video, and then merges them together to study some union features. To further
utilize the feature maps, two architectures are designed by using the
strategies of multi-scale and upsampling. Comparative experiments on iLIDS-VID,
PRID-2011 and MARS datasets illustrate that the proposed architectures are
significantly better for feature extraction than the state-of-the-art models.
|
Context Augmentation for Convolutional Neural Networks | Recent enhancements of deep convolutional neural networks (ConvNets)
empowered by enormous amounts of labeled data have closed the gap with human
performance for many object recognition tasks. These impressive results have
generated interest in understanding and visualization of ConvNets. In this
work, we study the effect of background in the task of image classification.
Our results show that changing the backgrounds of the training datasets can
have drastic effects on testing accuracies. Furthermore, we enhance existing
augmentation techniques with the foreground segmented objects. The findings of
this work are important in increasing the accuracies when only a small dataset
is available, in creating datasets, and creating synthetic images.
|
A Pitfall of Unsupervised Pre-Training | The point of this paper is to question typical assumptions in deep learning
and suggest alternatives. A particular contribution is to prove that even if a
Stacked Convolutional Auto-Encoder is good at reconstructing pictures, it is
not necessarily good at discriminating their classes. When using Auto-Encoders,
intuitively one assumes that features which are good for reconstruction will
also lead to high classification accuracy. Indeed, it became research practice
and is a suggested strategy by introductory books. However, we prove that this
is not always the case. We thoroughly investigate the quality of features
produced by Stacked Convolutional Auto-Encoders when trained to reconstruct
their input. In particular, we analyze the relation between the reconstruction
and classification capabilities of the network, if we were to use the same
features for both tasks. Experimental results suggest that in fact, there is no
correlation between the reconstruction score and the quality of features for a
classification task. This means, more formally, that the sub-dimension
representation space learned from the Stacked Convolutional Auto-Encoder (while
being trained for input reconstruction) is not necessarily better separable
than the initial input space. Furthermore, we show that the reconstruction
error is not a good metric to assess the quality of features, because it is
biased by the decoder quality. We do not question the usefulness of
pre-training, but we conclude that aiming for the lowest reconstruction error
is not necessarily a good idea if afterwards one performs a classification
task.
|
Learning a Generative Model for Validity in Complex Discrete Structures | Deep generative models have been successfully used to learn representations
for high-dimensional discrete spaces by representing discrete objects as
sequences and employing powerful sequence-based deep models. Unfortunately,
these sequence-based models often produce invalid sequences: sequences which do
not represent any underlying discrete structure; invalid sequences hinder the
utility of such models. As a step towards solving this problem, we propose to
learn a deep recurrent validator model, which can estimate whether a partial
sequence can function as the beginning of a full, valid sequence. This
validator provides insight as to how individual sequence elements influence the
validity of the overall sequence, and can be used to constrain sequence based
models to generate valid sequences -- and thus faithfully model discrete
objects. Our approach is inspired by reinforcement learning, where an oracle
which can evaluate validity of complete sequences provides a sparse reward
signal. We demonstrate its effectiveness as a generative model of Python 3
source code for mathematical expressions, and in improving the ability of a
variational autoencoder trained on SMILES strings to decode valid molecular
structures.
|
Differentially Private Dropout | Large data collections required for the training of neural networks often
contain sensitive information such as the medical histories of patients, and
the privacy of the training data must be preserved. In this paper, we introduce
a dropout technique that provides an elegant Bayesian interpretation to
dropout, and show that the intrinsic noise added, with the primary goal of
regularization, can be exploited to obtain a degree of differential privacy.
The iterative nature of training neural networks presents a challenge for
privacy-preserving estimation since multiple iterations increase the amount of
noise added. We overcome this by using a relaxed notion of differential
privacy, called concentrated differential privacy, which provides tighter
estimates on the overall privacy loss. We demonstrate the accuracy of our
privacy-preserving dropout algorithm on benchmark datasets.
|
An Ensemble of Deep Convolutional Neural Networks for Alzheimer's
Disease Detection and Classification | Alzheimer's Disease destroys brain cells causing people to lose their memory,
mental functions and ability to continue daily activities. It is a severe
neurological brain disorder which is not curable, but earlier detection of
Alzheimer's Disease can help for proper treatment and to prevent brain tissue
damage. Detection and classification of Alzheimer's Disease (AD) is challenging
because sometimes the signs that distinguish Alzheimer's Disease MRI data can
be found in normal healthy brain MRI data of older people. Moreover, there are
relatively small amount of dataset available to train the automated Alzheimer's
Disease detection and classification model. In this paper, we present a novel
Alzheimer's Disease detection and classification model using brain MRI data
analysis. We develop an ensemble of deep convolutional neural networks and
demonstrate superior performance on the Open Access Series of Imaging Studies
(OASIS) dataset.
|
Linear Convergence of An Iterative Phase Retrieval Algorithm with Data
Reuse | Phase retrieval has been an attractive but difficult problem rising from
physical science, and there has been a gap between state-of-the-art theoretical
convergence analyses and the corresponding efficient retrieval methods.
Firstly, these analyses all assume that the sensing vectors and the iterative
updates are independent, which only fits the ideal model with infinite
measurements but not the reality, where data are limited and have to be reused.
Secondly, the empirical results of some efficient methods, such as the
randomized Kaczmarz method, show linear convergence, which is beyond existing
theoretical explanations considering its randomness and reuse of data. In this
work, we study for the first time, without the independence assumption, the
convergence behavior of the randomized Kaczmarz method for phase retrieval.
Specifically, beginning from taking expectation of the squared estimation error
with respect to the index of measurement by fixing the sensing vector and the
error in the previous step, we discard the independence assumption, rigorously
derive the upper and lower bounds of the reduction of the mean squared error,
and prove the linear convergence. This work fills the gap between a fast
converging algorithm and its theoretical understanding. The proposed
methodology may contribute to the study of other iterative algorithms for phase
retrieval and other problems in the broad area of signal processing and machine
learning.
|
OL\'E: Orthogonal Low-rank Embedding, A Plug and Play Geometric Loss for
Deep Learning | Deep neural networks trained using a softmax layer at the top and the
cross-entropy loss are ubiquitous tools for image classification. Yet, this
does not naturally enforce intra-class similarity nor inter-class margin of the
learned deep representations. To simultaneously achieve these two goals,
different solutions have been proposed in the literature, such as the pairwise
or triplet losses. However, such solutions carry the extra task of selecting
pairs or triplets, and the extra computational burden of computing and learning
for many combinations of them. In this paper, we propose a plug-and-play loss
term for deep networks that explicitly reduces intra-class variance and
enforces inter-class margin simultaneously, in a simple and elegant geometric
manner. For each class, the deep features are collapsed into a learned linear
subspace, or union of them, and inter-class subspaces are pushed to be as
orthogonal as possible. Our proposed Orthogonal Low-rank Embedding (OL\'E) does
not require carefully crafting pairs or triplets of samples for training, and
works standalone as a classification loss, being the first reported deep metric
learning framework of its kind. Because of the improved margin between features
of different classes, the resulting deep networks generalize better, are more
discriminative, and more robust. We demonstrate improved classification
performance in general object recognition, plugging the proposed loss term into
existing off-the-shelf architectures. In particular, we show the advantage of
the proposed loss in the small data/model scenario, and we significantly
advance the state-of-the-art on the Stanford STL-10 benchmark.
|
Mastering Chess and Shogi by Self-Play with a General Reinforcement
Learning Algorithm | The game of chess is the most widely-studied domain in the history of
artificial intelligence. The strongest programs are based on a combination of
sophisticated search techniques, domain-specific adaptations, and handcrafted
evaluation functions that have been refined by human experts over several
decades. In contrast, the AlphaGo Zero program recently achieved superhuman
performance in the game of Go, by tabula rasa reinforcement learning from games
of self-play. In this paper, we generalise this approach into a single
AlphaZero algorithm that can achieve, tabula rasa, superhuman performance in
many challenging domains. Starting from random play, and given no domain
knowledge except the game rules, AlphaZero achieved within 24 hours a
superhuman level of play in the games of chess and shogi (Japanese chess) as
well as Go, and convincingly defeated a world-champion program in each case.
|
Towards Recovery of Conditional Vectors from Conditional Generative
Adversarial Networks | A conditional Generative Adversarial Network allows for generating samples
conditioned on certain external information. Being able to recover latent and
conditional vectors from a condi- tional GAN can be potentially valuable in
various applications, ranging from image manipulation for entertaining purposes
to diagnosis of the neural networks for security purposes. In this work, we
show that it is possible to recover both latent and conditional vectors from
generated images given the generator of a conditional generative adversarial
network. Such a recovery is not trivial due to the often multi-layered
non-linearity of deep neural networks. Furthermore, the effect of such recovery
applied on real natural images are investigated. We discovered that there
exists a gap between the recovery performance on generated and real images,
which we believe comes from the difference between generated data distribution
and real data distribution. Experiments are conducted to evaluate the recovered
conditional vectors and the reconstructed images from these recovered vectors
quantitatively and qualitatively, showing promising results.
|
Sum of previous inpatient serum creatinine measurements predicts acute
kidney injury in rehospitalized patients | Acute Kidney Injury (AKI), the abrupt decline in kidney function due to
temporary or permanent injury, is associated with increased mortality,
morbidity, length of stay, and hospital cost. Sometimes, simple interventions
such as medication review or hydration can prevent AKI. There is therefore
interest in estimating risk of AKI at hospitalization. To gain insight into
this task, we employ multilayer perceptron (MLP) and recurrent neural networks
(RNNs) using serum creatinine (sCr) as a lone feature. We explore different
feature input structures, including variable-length look-backs and a nested
formulation for rehospitalized patients with previous sCr measurements.
Experimental results show that the simplest model, MLP processing the sum of
sCr, had best performance: AUROC 0.92 and AUPRC 0.70. Such a simple model could
be easily integrated into an EHR. Preliminary results also suggest that
inpatient data streams with missing outpatient measurements---common in the
medical setting---might be best modeled with a tailored architecture.
|
Deep Gradient Compression: Reducing the Communication Bandwidth for
Distributed Training | Large-scale distributed training requires significant communication bandwidth
for gradient exchange that limits the scalability of multi-node training, and
requires expensive high-bandwidth network infrastructure. The situation gets
even worse with distributed training on mobile devices (federated learning),
which suffers from higher latency, lower throughput, and intermittent poor
connections. In this paper, we find 99.9% of the gradient exchange in
distributed SGD is redundant, and propose Deep Gradient Compression (DGC) to
greatly reduce the communication bandwidth. To preserve accuracy during
compression, DGC employs four methods: momentum correction, local gradient
clipping, momentum factor masking, and warm-up training. We have applied Deep
Gradient Compression to image classification, speech recognition, and language
modeling with multiple datasets including Cifar10, ImageNet, Penn Treebank, and
Librispeech Corpus. On these scenarios, Deep Gradient Compression achieves a
gradient compression ratio from 270x to 600x without losing accuracy, cutting
the gradient size of ResNet-50 from 97MB to 0.35MB, and for DeepSpeech from
488MB to 0.74MB. Deep gradient compression enables large-scale distributed
training on inexpensive commodity 1Gbps Ethernet and facilitates distributed
training on mobile. Code is available at:
https://github.com/synxlin/deep-gradient-compression.
|
Online Learning with Gated Linear Networks | This paper describes a family of probabilistic architectures designed for
online learning under the logarithmic loss. Rather than relying on non-linear
transfer functions, our method gains representational power by the use of data
conditioning. We state under general conditions a learnable capacity theorem
that shows this approach can in principle learn any bounded Borel-measurable
function on a compact subset of euclidean space; the result is stronger than
many universality results for connectionist architectures because we provide
both the model and the learning procedure for which convergence is guaranteed.
|
How to Learn a Model Checker | We show how machine-learning techniques, particularly neural networks, offer
a very effective and highly efficient solution to the approximate
model-checking problem for continuous and hybrid systems, a solution where the
general-purpose model checker is replaced by a model-specific classifier
trained by sampling model trajectories. To the best of our knowledge, we are
the first to establish this link from machine learning to model checking. Our
method comprises a pipeline of analysis techniques for estimating and obtaining
statistical guarantees on the classifier's prediction performance, as well as
tuning techniques to improve such performance. Our experimental evaluation
considers the time-bounded reachability problem for three well-established
benchmarks in the hybrid systems community. On these examples, we achieve an
accuracy of 99.82% to 100% and a false-negative rate (incorrectly predicting
that unsafe states are not reachable from a given state) of 0.0007 to 0. We
believe that this level of accuracy is acceptable in many practical
applications and we show how the approximate model checker can be made more
conservative by tuning the classifier through further training and selection of
the classification threshold.
|
Regularization and feature selection for large dimensional data | Feature selection has evolved to be an important step in several machine
learning paradigms. In domains like bio-informatics and text classification
which involve data of high dimensions, feature selection can help in
drastically reducing the feature space. In cases where it is difficult or
infeasible to obtain sufficient number of training examples, feature selection
helps overcome the curse of dimensionality which in turn helps improve
performance of the classification algorithm. The focus of our research here are
five embedded feature selection methods which use either the ridge regression,
or Lasso regression, or a combination of the two in the regularization part of
the optimization function. We evaluate five chosen methods on five large
dimensional datasets and compare them on the parameters of sparsity and
correlation in the datasets and their execution times.
|
Single-trial P300 Classification using PCA with LDA, QDA and Neural
Networks | The P300 event-related potential (ERP), evoked in scalp-recorded
electroencephalography (EEG) by external stimuli, has proven to be a reliable
response for controlling a BCI. The P300 component of an event related
potential is thus widely used in brain-computer interfaces to translate the
subjects' intent by mere thoughts into commands to control artificial devices.
The main challenge in the classification of P300 trials in
electroencephalographic (EEG) data is the low signal-to-noise ratio (SNR) of
the P300 response. To overcome the low SNR of individual trials, it is common
practice to average together many consecutive trials, which effectively
diminishes the random noise. Unfortunately, when more repeated trials are
required for applications such as the P300 speller, the communication rate is
greatly reduced. This has resulted in a need for better methods to improve
single-trial classification accuracy of P300 response. In this work, we use
Principal Component Analysis (PCA) as a preprocessing method and use Linear
Discriminant Analysis (LDA)and neural networks for classification. The results
show that a combination of PCA with these methods provided as high as 13\%
accuracy gain for single-trial classification while using only 3 to 4 principal
components.
|
A Scalable Deep Neural Network Architecture for Multi-Building and
Multi-Floor Indoor Localization Based on Wi-Fi Fingerprinting | One of the key technologies for future large-scale location-aware services
covering a complex of multi-story buildings --- e.g., a big shopping mall and a
university campus --- is a scalable indoor localization technique. In this
paper, we report the current status of our investigation on the use of deep
neural networks (DNNs) for scalable building/floor classification and
floor-level position estimation based on Wi-Fi fingerprinting. Exploiting the
hierarchical nature of the building/floor estimation and floor-level
coordinates estimation of a location, we propose a new DNN architecture
consisting of a stacked autoencoder for the reduction of feature space
dimension and a feed-forward classifier for multi-label classification of
building/floor/location, on which the multi-building and multi-floor indoor
localization system based on Wi-Fi fingerprinting is built. Experimental
results for the performance of building/floor estimation and floor-level
coordinates estimation of a given location demonstrate the feasibility of the
proposed DNN-based indoor localization system, which can provide near
state-of-the-art performance using a single DNN, for the implementation with
lower complexity and energy consumption at mobile devices.
|
AdaBatch: Adaptive Batch Sizes for Training Deep Neural Networks | Training deep neural networks with Stochastic Gradient Descent, or its
variants, requires careful choice of both learning rate and batch size. While
smaller batch sizes generally converge in fewer training epochs, larger batch
sizes offer more parallelism and hence better computational efficiency. We have
developed a new training approach that, rather than statically choosing a
single batch size for all epochs, adaptively increases the batch size during
the training process. Our method delivers the convergence rate of small batch
sizes while achieving performance similar to large batch sizes. We analyse our
approach using the standard AlexNet, ResNet, and VGG networks operating on the
popular CIFAR-10, CIFAR-100, and ImageNet datasets. Our results demonstrate
that learning with adaptive batch sizes can improve performance by factors of
up to 6.25 on 4 NVIDIA Tesla P100 GPUs while changing accuracy by less than 1%
relative to training with fixed batch sizes.
|
SMILES2Vec: An Interpretable General-Purpose Deep Neural Network for
Predicting Chemical Properties | Chemical databases store information in text representations, and the SMILES
format is a universal standard used in many cheminformatics software. Encoded
in each SMILES string is structural information that can be used to predict
complex chemical properties. In this work, we develop SMILES2vec, a deep RNN
that automatically learns features from SMILES to predict chemical properties,
without the need for additional explicit feature engineering. Using Bayesian
optimization methods to tune the network architecture, we show that an
optimized SMILES2vec model can serve as a general-purpose neural network for
predicting distinct chemical properties including toxicity, activity,
solubility and solvation energy, while also outperforming contemporary MLP
neural networks that uses engineered features. Furthermore, we demonstrate
proof-of-concept of interpretability by developing an explanation mask that
localizes on the most important characters used in making a prediction. When
tested on the solubility dataset, it identified specific parts of a chemical
that is consistent with established first-principles knowledge with an accuracy
of 88%. Our work demonstrates that neural networks can learn technically
accurate chemical concept and provide state-of-the-art accuracy, making
interpretable deep neural networks a useful tool of relevance to the chemical
industry.
|
Bayesian Policy Gradients via Alpha Divergence Dropout Inference | Policy gradient methods have had great success in solving continuous control
tasks, yet the stochastic nature of such problems makes deterministic value
estimation difficult. We propose an approach which instead estimates a
distribution by fitting the value function with a Bayesian Neural Network. We
optimize an $\alpha$-divergence objective with Bayesian dropout approximation
to learn and estimate this distribution. We show that using the Monte Carlo
posterior mean of the Bayesian value function distribution, rather than a
deterministic network, improves stability and performance of policy gradient
methods in continuous control MuJoCo simulations.
|
Learning General Latent-Variable Graphical Models with Predictive Belief
Propagation | Learning general latent-variable probabilistic graphical models is a key
theoretical challenge in machine learning and artificial intelligence. All
previous methods, including the EM algorithm and the spectral algorithms, face
severe limitations that largely restrict their applicability and affect their
performance. In order to overcome these limitations, in this paper we introduce
a novel formulation of message-passing inference over junction trees named
predictive belief propagation, and propose a new learning and inference
algorithm for general latent-variable graphical models based on this
formulation. Our proposed algorithm reduces the hard parameter learning problem
into a sequence of supervised learning problems, and unifies the learning of
different kinds of latent graphical models into a single learning framework,
which is local-optima-free and statistically consistent. We then give a proof
of the correctness of our algorithm and show in experiments on both synthetic
and real datasets that our algorithm significantly outperforms both the EM
algorithm and the spectral algorithm while also being orders of magnitude
faster to compute.
|
Enabling Early Audio Event Detection with Neural Networks | This paper presents a methodology for early detection of audio events from
audio streams. Early detection is the ability to infer an ongoing event during
its initial stage. The proposed system consists of a novel inference step
coupled with dual parallel tailored-loss deep neural networks (DNNs). The DNNs
share a similar architecture except for their loss functions, i.e. weighted
loss and multitask loss, which are designed to efficiently cope with issues
common to audio event detection. The inference step is newly introduced to make
use of the network outputs for recognizing ongoing events. The monotonicity of
the detection function is required for reliable early detection, and will also
be proved. Experiments on the ITC-Irst database show that the proposed system
achieves state-of-the-art detection performance. Furthermore, even partial
events are sufficient to achieve good performance similar to that obtained when
an entire event is observed, enabling early event detection.
|
CNN training with graph-based sample preselection: application to
handwritten character recognition | In this paper, we present a study on sample preselection in large training
data set for CNN-based classification. To do so, we structure the input data
set in a network representation, namely the Relative Neighbourhood Graph, and
then extract some vectors of interest. The proposed preselection method is
evaluated in the context of handwritten character recognition, by using two
data sets, up to several hundred thousands of images. It is shown that the
graph-based preselection can reduce the training data set without degrading the
recognition accuracy of a non pretrained CNN shallow model.
|
Listening to Chaotic Whispers: A Deep Learning Framework for
News-oriented Stock Trend Prediction | Stock trend prediction plays a critical role in seeking maximized profit from
stock investment. However, precise trend prediction is very difficult since the
highly volatile and non-stationary nature of stock market. Exploding
information on Internet together with advancing development of natural language
processing and text mining techniques have enable investors to unveil market
trends and volatility from online content. Unfortunately, the quality,
trustworthiness and comprehensiveness of online content related to stock market
varies drastically, and a large portion consists of the low-quality news,
comments, or even rumors. To address this challenge, we imitate the learning
process of human beings facing such chaotic online news, driven by three
principles: sequential content dependency, diverse influence, and effective and
efficient learning. In this paper, to capture the first two principles, we
designed a Hybrid Attention Networks to predict the stock trend based on the
sequence of recent related news. Moreover, we apply the self-paced learning
mechanism to imitate the third principle. Extensive experiments on real-world
stock market data demonstrate the effectiveness of our approach.
|
Guided Labeling using Convolutional Neural Networks | Over the last couple of years, deep learning and especially convolutional
neural networks have become one of the work horses of computer vision. One
limiting factor for the applicability of supervised deep learning to more areas
is the need for large, manually labeled datasets. In this paper we propose an
easy to implement method we call guided labeling, which automatically
determines which samples from an unlabeled dataset should be labeled. We show
that using this procedure, the amount of samples that need to be labeled is
reduced considerably in comparison to labeling images arbitrarily.
|
Distribution-Based Categorization of Classifier Transfer Learning | Transfer Learning (TL) aims to transfer knowledge acquired in one problem,
the source problem, onto another problem, the target problem, dispensing with
the bottom-up construction of the target model. Due to its relevance, TL has
gained significant interest in the Machine Learning community since it paves
the way to devise intelligent learning models that can easily be tailored to
many different applications. As it is natural in a fast evolving area, a wide
variety of TL methods, settings and nomenclature have been proposed so far.
However, a wide range of works have been reporting different names for the same
concepts. This concept and terminology mixture contribute however to obscure
the TL field, hindering its proper consideration. In this paper we present a
review of the literature on the majority of classification TL methods, and also
a distribution-based categorization of TL with a common nomenclature suitable
to classification problems. Under this perspective three main TL categories are
presented, discussed and illustrated with examples.
|
A trans-disciplinary review of deep learning research for water
resources scientists | Deep learning (DL), a new-generation of artificial neural network research,
has transformed industries, daily lives and various scientific disciplines in
recent years. DL represents significant progress in the ability of neural
networks to automatically engineer problem-relevant features and capture highly
complex data distributions. I argue that DL can help address several major new
and old challenges facing research in water sciences such as
inter-disciplinarity, data discoverability, hydrologic scaling, equifinality,
and needs for parameter regionalization. This review paper is intended to
provide water resources scientists and hydrologists in particular with a simple
technical overview, trans-disciplinary progress update, and a source of
inspiration about the relevance of DL to water. The review reveals that various
physical and geoscientific disciplines have utilized DL to address data
challenges, improve efficiency, and gain scientific insights. DL is especially
suited for information extraction from image-like data and sequential data.
Techniques and experiences presented in other disciplines are of high relevance
to water research. Meanwhile, less noticed is that DL may also serve as a
scientific exploratory tool. A new area termed 'AI neuroscience,' where
scientists interpret the decision process of deep networks and derive insights,
has been born. This budding sub-discipline has demonstrated methods including
correlation-based analysis, inversion of network-extracted features,
reduced-order approximations by interpretable models, and attribution of
network decisions to inputs. Moreover, DL can also use data to condition
neurons that mimic problem-specific fundamental organizing units, thus
revealing emergent behaviors of these units. Vast opportunities exist for DL to
propel advances in water sciences.
|
LocNet: Global localization in 3D point clouds for mobile vehicles | Global localization in 3D point clouds is a challenging problem of estimating
the pose of vehicles without any prior knowledge. In this paper, a solution to
this problem is presented by achieving place recognition and metric pose
estimation in the global prior map. Specifically, we present a semi-handcrafted
representation learning method for LiDAR point clouds using siamese LocNets,
which states the place recognition problem to a similarity modeling problem.
With the final learned representations by LocNet, a global localization
framework with range-only observations is proposed. To demonstrate the
performance and effectiveness of our global localization system, KITTI dataset
is employed for comparison with other algorithms, and also on our long-time
multi-session datasets for evaluation. The result shows that our system can
achieve high accuracy.
|
Lung Nodule Classification by the Combination of Fusion Classifier and
Cascaded Convolutional Neural Networks | Lung nodule classification is a class imbalanced problem, as nodules are
found with much lower frequency than non-nodules. In the class imbalanced
problem, conventional classifiers tend to be overwhelmed by the majority class
and ignore the minority class. We showed that cascaded convolutional neural
networks can classify the nodule candidates precisely for a class imbalanced
nodule candidate data set in our previous study. In this paper, we propose
Fusion classifier in conjunction with the cascaded convolutional neural network
models. To fuse the models, nodule probabilities are calculated by using the
convolutional neural network models at first. Then, Fusion classifier is
trained and tested by the nodule probabilities. The proposed method achieved
the sensitivity of 94.4% and 95.9% at 4 and 8 false positives per scan in Free
Receiver Operating Characteristics (FROC) curve analysis, respectively.
|
Why Do Neural Dialog Systems Generate Short and Meaningless Replies? A
Comparison between Dialog and Translation | This paper addresses the question: Why do neural dialog systems generate
short and meaningless replies? We conjecture that, in a dialog system, an
utterance may have multiple equally plausible replies, causing the deficiency
of neural networks in the dialog application. We propose a systematic way to
mimic the dialog scenario in a machine translation system, and manage to
reproduce the phenomenon of generating short and less meaningful sentences in
the translation setting, showing evidence of our conjecture.
|
Attention based convolutional neural network for predicting RNA-protein
binding sites | RNA-binding proteins (RBPs) play crucial roles in many biological processes,
e.g. gene regulation. Computational identification of RBP binding sites on RNAs
are urgently needed. In particular, RBPs bind to RNAs by recognizing sequence
motifs. Thus, fast locating those motifs on RNA sequences is crucial and
time-efficient for determining whether the RNAs interact with the RBPs or not.
In this study, we present an attention based convolutional neural network,
iDeepA, to predict RNA-protein binding sites from raw RNA sequences. We first
encode RNA sequences into one-hot encoding. Next, we design a deep learning
model with a convolutional neural network (CNN) and an attention mechanism,
which automatically search for important positions, e.g. binding motifs, to
learn discriminant high-level features for predicting RBP binding sites. We
evaluate iDeepA on publicly gold-standard RBP binding sites derived from
CLIP-seq data. The results demonstrate iDeepA achieves comparable performance
with other state-of-the-art methods.
|
Generative Adversarial Perturbations | In this paper, we propose novel generative models for creating adversarial
examples, slightly perturbed images resembling natural images but maliciously
crafted to fool pre-trained models. We present trainable deep neural networks
for transforming images to adversarial perturbations. Our proposed models can
produce image-agnostic and image-dependent perturbations for both targeted and
non-targeted attacks. We also demonstrate that similar architectures can
achieve impressive results in fooling classification and semantic segmentation
models, obviating the need for hand-crafting attack methods for each task.
Using extensive experiments on challenging high-resolution datasets such as
ImageNet and Cityscapes, we show that our perturbations achieve high fooling
rates with small perturbation norms. Moreover, our attacks are considerably
faster than current iterative methods at inference time.
|
SGAN: An Alternative Training of Generative Adversarial Networks | The Generative Adversarial Networks (GANs) have demonstrated impressive
performance for data synthesis, and are now used in a wide range of computer
vision tasks. In spite of this success, they gained a reputation for being
difficult to train, what results in a time-consuming and human-involved
development process to use them.
We consider an alternative training process, named SGAN, in which several
adversarial "local" pairs of networks are trained independently so that a
"global" supervising pair of networks can be trained against them. The goal is
to train the global pair with the corresponding ensemble opponent for improved
performances in terms of mode coverage. This approach aims at increasing the
chances that learning will not stop for the global pair, preventing both to be
trapped in an unsatisfactory local minimum, or to face oscillations often
observed in practice. To guarantee the latter, the global pair never affects
the local ones.
The rules of SGAN training are thus as follows: the global generator and
discriminator are trained using the local discriminators and generators,
respectively, whereas the local networks are trained with their fixed local
opponent.
Experimental results on both toy and real-world problems demonstrate that
this approach outperforms standard training in terms of better mitigating mode
collapse, stability while converging and that it surprisingly, increases the
convergence speed as well.
|
Noisy Natural Gradient as Variational Inference | Variational Bayesian neural nets combine the flexibility of deep learning
with Bayesian uncertainty estimation. Unfortunately, there is a tradeoff
between cheap but simple variational families (e.g.~fully factorized) or
expensive and complicated inference procedures. We show that natural gradient
ascent with adaptive weight noise implicitly fits a variational posterior to
maximize the evidence lower bound (ELBO). This insight allows us to train
full-covariance, fully factorized, or matrix-variate Gaussian variational
posteriors using noisy versions of natural gradient, Adam, and K-FAC,
respectively, making it possible to scale up to modern-size ConvNets. On
standard regression benchmarks, our noisy K-FAC algorithm makes better
predictions and matches Hamiltonian Monte Carlo's predictive variances better
than existing methods. Its improved uncertainty estimates lead to more
efficient exploration in active learning, and intrinsic motivation for
reinforcement learning.
|
High performance ultra-low-precision convolutions on mobile devices | Many applications of mobile deep learning, especially real-time computer
vision workloads, are constrained by computation power. This is particularly
true for workloads running on older consumer phones, where a typical device
might be powered by a single- or dual-core ARMv7 CPU. We provide an open-source
implementation and a comprehensive analysis of (to our knowledge) the state of
the art ultra-low-precision (<4 bit precision) implementation of the core
primitives required for modern deep learning workloads on ARMv7 devices, and
demonstrate speedups of 4x-20x over our additional state-of-the-art float32 and
int8 baselines.
|
HyperPower: Power- and Memory-Constrained Hyper-Parameter Optimization
for Neural Networks | While selecting the hyper-parameters of Neural Networks (NNs) has been so far
treated as an art, the emergence of more complex, deeper architectures poses
increasingly more challenges to designers and Machine Learning (ML)
practitioners, especially when power and memory constraints need to be
considered. In this work, we propose HyperPower, a framework that enables
efficient Bayesian optimization and random search in the context of power- and
memory-constrained hyper-parameter optimization for NNs running on a given
hardware platform. HyperPower is the first work (i) to show that power
consumption can be used as a low-cost, a priori known constraint, and (ii) to
propose predictive models for the power and memory of NNs executing on GPUs.
Thanks to HyperPower, the number of function evaluations and the best test
error achieved by a constraint-unaware method are reached up to 112.99x and
30.12x faster, respectively, while never considering invalid configurations.
HyperPower significantly speeds up the hyper-parameter optimization, achieving
up to 57.20x more function evaluations compared to constraint-unaware methods
for a given time interval, effectively yielding significant accuracy
improvements by up to 67.6%.
|
Deep Primal-Dual Reinforcement Learning: Accelerating Actor-Critic using
Bellman Duality | We develop a parameterized Primal-Dual $\pi$ Learning method based on deep
neural networks for Markov decision process with large state space and
off-policy reinforcement learning. In contrast to the popular Q-learning and
actor-critic methods that are based on successive approximations to the
nonlinear Bellman equation, our method makes primal-dual updates to the policy
and value functions utilizing the fundamental linear Bellman duality. Naive
parametrization of the primal-dual $\pi$ learning method using deep neural
networks would encounter two major challenges: (1) each update requires
computing a probability distribution over the state space and is intractable;
(2) the iterates are unstable since the parameterized Lagrangian function is no
longer linear. We address these challenges by proposing a relaxed Lagrangian
formulation with a regularization penalty using the advantage function. We show
that the dual policy update step in our method is equivalent to the policy
gradient update in the actor-critic method in some special case, while the
value updates differ substantially. The main advantage of the primal-dual $\pi$
learning method lies in that the value and policy updates are closely coupled
together using the Bellman duality and therefore more informative. Experiments
on a simple cart-pole problem show that the algorithm significantly outperforms
the one-step temporal-difference actor-critic method, which is the most
relevant benchmark method to compare with. We believe that the primal-dual
updates to the value and policy functions would expedite the learning process.
The proposed methods might open a door to more efficient algorithms and sharper
theoretical analysis.
|
Cost-sensitive detection with variational autoencoders for environmental
acoustic sensing | Environmental acoustic sensing involves the retrieval and processing of audio
signals to better understand our surroundings. While large-scale acoustic data
make manual analysis infeasible, they provide a suitable playground for machine
learning approaches. Most existing machine learning techniques developed for
environmental acoustic sensing do not provide flexible control of the trade-off
between the false positive rate and the false negative rate. This paper
presents a cost-sensitive classification paradigm, in which the
hyper-parameters of classifiers and the structure of variational autoencoders
are selected in a principled Neyman-Pearson framework. We examine the
performance of the proposed approach using a dataset from the HumBug project
which aims to detect the presence of mosquitoes using sound collected by simple
embedded devices.
|
Adversarial Examples that Fool Detectors | An adversarial example is an example that has been adjusted to produce a
wrong label when presented to a system at test time. To date, adversarial
example constructions have been demonstrated for classifiers, but not for
detectors. If adversarial examples that could fool a detector exist, they could
be used to (for example) maliciously create security hazards on roads populated
with smart vehicles. In this paper, we demonstrate a construction that
successfully fools two standard detectors, Faster RCNN and YOLO. The existence
of such examples is surprising, as attacking a classifier is very different
from attacking a detector, and that the structure of detectors - which must
search for their own bounding box, and which cannot estimate that box very
accurately - makes it quite likely that adversarial patterns are strongly
disrupted. We show that our construction produces adversarial examples that
generalize well across sequences digitally, even though large perturbations are
needed. We also show that our construction yields physical objects that are
adversarial.
|
CNNs are Globally Optimal Given Multi-Layer Support | Stochastic Gradient Descent (SGD) is the central workhorse for training
modern CNNs. Although giving impressive empirical performance it can be slow to
converge. In this paper we explore a novel strategy for training a CNN using an
alternation strategy that offers substantial speedups during training. We make
the following contributions: (i) replace the ReLU non-linearity within a CNN
with positive hard-thresholding, (ii) reinterpret this non-linearity as a
binary state vector making the entire CNN linear if the multi-layer support is
known, and (iii) demonstrate that under certain conditions a global optima to
the CNN can be found through local descent. We then employ a novel alternation
strategy (between weights and support) for CNN training that leads to
substantially faster convergence rates, nice theoretical properties, and
achieving state of the art results across large scale datasets (e.g. ImageNet)
as well as other standard benchmarks.
|
Take it in your stride: Do we need striding in CNNs? | Since their inception, CNNs have utilized some type of striding operator to
reduce the overlap of receptive fields and spatial dimensions. Although having
clear heuristic motivations (i.e. lowering the number of parameters to learn)
the mathematical role of striding within CNN learning remains unclear. This
paper offers a novel and mathematical rigorous perspective on the role of the
striding operator within modern CNNs. Specifically, we demonstrate
theoretically that one can always represent a CNN that incorporates striding
with an equivalent non-striding CNN which has more filters and smaller size.
Through this equivalence we are then able to characterize striding as an
additional mechanism for parameter sharing among channels, thus reducing
training complexity. Finally, the framework presented in this paper offers a
new mathematical perspective on the role of striding which we hope shall
facilitate and simplify the future theoretical analysis of CNNs.
|
Semi-Supervised Learning with IPM-based GANs: an Empirical Study | We present an empirical investigation of a recent class of Generative
Adversarial Networks (GANs) using Integral Probability Metrics (IPM) and their
performance for semi-supervised learning. IPM-based GANs like Wasserstein GAN,
Fisher GAN and Sobolev GAN have desirable properties in terms of theoretical
understanding, training stability, and a meaningful loss. In this work we
investigate how the design of the critic (or discriminator) influences the
performance in semi-supervised learning. We distill three key take-aways which
are important for good SSL performance: (1) the K+1 formulation, (2) avoiding
batch normalization in the critic and (3) avoiding gradient penalty constraints
on the classification layer.
|
Gini-regularized Optimal Transport with an Application to
Spatio-Temporal Forecasting | Rapidly growing product lines and services require a finer-granularity
forecast that considers geographic locales. However the open question remains,
how to assess the quality of a spatio-temporal forecast? In this manuscript we
introduce a metric to evaluate spatio-temporal forecasts. This metric is based
on an Opti- mal Transport (OT) problem. The metric we propose is a constrained
OT objec- tive function using the Gini impurity function as a regularizer. We
demonstrate through computer experiments both the qualitative and the
quantitative charac- teristics of the Gini regularized OT problem. Moreover, we
show that the Gini regularized OT problem converges to the classical OT
problem, when the Gini regularized problem is considered as a function of
{\lambda}, the regularization parame-ter. The convergence to the classical OT
solution is faster than the state-of-the-art Entropic-regularized OT[Cuturi,
2013] and results in a numerically more stable algorithm.
|
Learning Random Fourier Features by Hybrid Constrained Optimization | The kernel embedding algorithm is an important component for adapting kernel
methods to large datasets. Since the algorithm consumes a major computation
cost in the testing phase, we propose a novel teacher-learner framework of
learning computation-efficient kernel embeddings from specific data. In the
framework, the high-precision embeddings (teacher) transfer the data
information to the computation-efficient kernel embeddings (learner). We
jointly select informative embedding functions and pursue an orthogonal
transformation between two embeddings. We propose a novel approach of
constrained variational expectation maximization (CVEM), where the alternate
direction method of multiplier (ADMM) is applied over a nonconvex domain in the
maximization step. We also propose two specific formulations based on the
prevalent Random Fourier Feature (RFF), the masked and blocked version of
Computation-Efficient RFF (CERF), by imposing a random binary mask or a block
structure on the transformation matrix. By empirical studies of several
applications on different real-world datasets, we demonstrate that the CERF
significantly improves the performance of kernel methods upon the RFF, under
certain arithmetic operation requirements, and suitable for structured matrix
multiplication in Fastfood type algorithms.
|
Solving internal covariate shift in deep learning with linked neurons | This work proposes a novel solution to the problem of internal covariate
shift and dying neurons using the concept of linked neurons. We define the
neuron linkage in terms of two constraints: first, all neuron activations in
the linkage must have the same operating point. That is to say, all of them
share input weights. Secondly, a set of neurons is linked if and only if there
is at least one member of the linkage that has a non-zero gradient in regard to
the input of the activation function. This means that for any input in the
activation function, there is at least one member of the linkage that operates
in a non-flat and non-zero area. This simple change has profound implications
in the network learning dynamics. In this article we explore the consequences
of this proposal and show that by using this kind of units, internal covariate
shift is implicitly solved. As a result of this, the use of linked neurons
allows to train arbitrarily large networks without any architectural or
algorithmic trick, effectively removing the need of using re-normalization
schemes such as Batch Normalization, which leads to halving the required
training time. It also solves the problem of the need for standarized input
data. Results show that the units using the linkage not only do effectively
solve the aforementioned problems, but are also a competitive alternative with
respect to state-of-the-art with very promising results.
|
Differentially Private Variational Dropout | Deep neural networks with their large number of parameters are highly
flexible learning systems. The high flexibility in such networks brings with
some serious problems such as overfitting, and regularization is used to
address this problem. A currently popular and effective regularization
technique for controlling the overfitting is dropout. Often, large data
collections required for neural networks contain sensitive information such as
the medical histories of patients, and the privacy of the training data should
be protected. In this paper, we modify the recently proposed variational
dropout technique which provided an elegant Bayesian interpretation to dropout,
and show that the intrinsic noise in the variational dropout can be exploited
to obtain a degree of differential privacy. The iterative nature of training
neural networks presents a challenge for privacy-preserving estimation since
multiple iterations increase the amount of noise added. We overcome this by
using a relaxed notion of differential privacy, called concentrated
differential privacy, which provides tighter estimates on the overall privacy
loss. We demonstrate the accuracy of our privacy-preserving variational dropout
algorithm on benchmark datasets.
|
How consistent is my model with the data? Information-Theoretic Model
Check | The choice of model class is fundamental in statistical learning and system
identification, no matter whether the class is derived from physical principles
or is a generic black-box. We develop a method to evaluate the specified model
class by assessing its capability of reproducing data that is similar to the
observed data record. This model check is based on the information-theoretic
properties of models viewed as data generators and is applicable to e.g.
sequential data and nonlinear dynamical models. The method can be understood as
a specific two-sided posterior predictive test. We apply the
information-theoretic model check to both synthetic and real data and compare
it with a classical whiteness test.
|
AdaComp : Adaptive Residual Gradient Compression for Data-Parallel
Distributed Training | Highly distributed training of Deep Neural Networks (DNNs) on future compute
platforms (offering 100 of TeraOps/s of computational capacity) is expected to
be severely communication constrained. To overcome this limitation, new
gradient compression techniques are needed that are computationally friendly,
applicable to a wide variety of layers seen in Deep Neural Networks and
adaptable to variations in network architectures as well as their
hyper-parameters. In this paper we introduce a novel technique - the Adaptive
Residual Gradient Compression (AdaComp) scheme. AdaComp is based on localized
selection of gradient residues and automatically tunes the compression rate
depending on local activity. We show excellent results on a wide spectrum of
state of the art Deep Learning models in multiple domains (vision, speech,
language), datasets (MNIST, CIFAR10, ImageNet, BN50, Shakespeare), optimizers
(SGD with momentum, Adam) and network parameters (number of learners,
minibatch-size etc.). Exploiting both sparsity and quantization, we demonstrate
end-to-end compression rates of ~200X for fully-connected and recurrent layers,
and ~40X for convolutional layers, without any noticeable degradation in model
accuracies.
|
Using Rule-Based Labels for Weak Supervised Learning: A ChemNet for
Transferable Chemical Property Prediction | With access to large datasets, deep neural networks (DNN) have achieved
human-level accuracy in image and speech recognition tasks. However, in
chemistry, data is inherently small and fragmented. In this work, we develop an
approach of using rule-based knowledge for training ChemNet, a transferable and
generalizable deep neural network for chemical property prediction that learns
in a weak-supervised manner from large unlabeled chemical databases. When
coupled with transfer learning approaches to predict other smaller datasets for
chemical properties that it was not originally trained on, we show that
ChemNet's accuracy outperforms contemporary DNN models that were trained using
conventional supervised learning. Furthermore, we demonstrate that the ChemNet
pre-training approach is equally effective on both CNN (Chemception) and RNN
(SMILES2vec) models, indicating that this approach is network architecture
agnostic and is effective across multiple data modalities. Our results indicate
a pre-trained ChemNet that incorporates chemistry domain knowledge, enables the
development of generalizable neural networks for more accurate prediction of
novel chemical properties.
|
End-to-end Learning of Deterministic Decision Trees | Conventional decision trees have a number of favorable properties, including
interpretability, a small computational footprint and the ability to learn from
little training data. However, they lack a key quality that has helped fuel the
deep learning revolution: that of being end-to-end trainable, and to learn from
scratch those features that best allow to solve a given supervised learning
problem. Recent work (Kontschieder 2015) has addressed this deficit, but at the
cost of losing a main attractive trait of decision trees: the fact that each
sample is routed along a small subset of tree nodes only. We here propose a
model and Expectation-Maximization training scheme for decision trees that are
fully probabilistic at train time, but after a deterministic annealing process
become deterministic at test time. We also analyze the learned oblique split
parameters on image datasets and show that Neural Networks can be trained at
each split node. In summary, we present the first end-to-end learning scheme
for deterministic decision trees and present results on par with or superior to
published standard oblique decision tree algorithms.
|
Topics and Label Propagation: Best of Both Worlds for Weakly Supervised
Text Classification | We propose a Label Propagation based algorithm for weakly supervised text
classification. We construct a graph where each document is represented by a
node and edge weights represent similarities among the documents. Additionally,
we discover underlying topics using Latent Dirichlet Allocation (LDA) and
enrich the document graph by including the topics in the form of additional
nodes. The edge weights between a topic and a text document represent level of
"affinity" between them. Our approach does not require document level
labelling, instead it expects manual labels only for topic nodes. This
significantly minimizes the level of supervision needed as only a few topics
are observed to be enough for achieving sufficiently high accuracy. The Label
Propagation Algorithm is employed on this enriched graph to propagate labels
among the nodes. Our approach combines the advantages of Label Propagation
(through document-document similarities) and Topic Modelling (for minimal but
smart supervision). We demonstrate the effectiveness of our approach on various
datasets and compare with state-of-the-art weakly supervised text
classification approaches.
|
Exploring the Landscape of Spatial Robustness | The study of adversarial robustness has so far largely focused on
perturbations bound in p-norms. However, state-of-the-art models turn out to be
also vulnerable to other, more natural classes of perturbations such as
translations and rotations. In this work, we thoroughly investigate the
vulnerability of neural network--based classifiers to rotations and
translations. While data augmentation offers relatively small robustness, we
use ideas from robust optimization and test-time input aggregation to
significantly improve robustness. Finally we find that, in contrast to the
p-norm case, first-order methods cannot reliably find worst-case perturbations.
This highlights spatial robustness as a fundamentally different setting
requiring additional study. Code available at
https://github.com/MadryLab/adversarial_spatial and
https://github.com/MadryLab/spatial-pytorch.
|
RelNN: A Deep Neural Model for Relational Learning | Statistical relational AI (StarAI) aims at reasoning and learning in noisy
domains described in terms of objects and relationships by combining
probability with first-order logic. With huge advances in deep learning in the
current years, combining deep networks with first-order logic has been the
focus of several recent studies. Many of the existing attempts, however, only
focus on relations and ignore object properties. The attempts that do consider
object properties are limited in terms of modelling power or scalability. In
this paper, we develop relational neural networks (RelNNs) by adding hidden
layers to relational logistic regression (the relational counterpart of
logistic regression). We learn latent properties for objects both directly and
through general rules. Back-propagation is used for training these models. A
modular, layer-wise architecture facilitates utilizing the techniques developed
within deep learning community to our architecture. Initial experiments on
eight tasks over three real-world datasets show that RelNNs are promising
models for relational learning.
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.