title
stringlengths 7
246
| abstract
stringlengths 6
3.31k
|
|---|---|
Minimax Lower Bounds for Kronecker-Structured Dictionary Learning | Dictionary learning is the problem of estimating the collection of atomic
elements that provide a sparse representation of measured/collected signals or
data. This paper finds fundamental limits on the sample complexity of
estimating dictionaries for tensor data by proving a lower bound on the minimax
risk. This lower bound depends on the dimensions of the tensor and parameters
of the generative model. The focus of this paper is on second-order tensor
data, with the underlying dictionaries constructed by taking the Kronecker
product of two smaller dictionaries and the observed data generated by sparse
linear combinations of dictionary atoms observed through white Gaussian noise.
In this regard, the paper provides a general lower bound on the minimax risk
and also adapts the proof techniques for equivalent results using sparse and
Gaussian coefficient models. The reported results suggest that the sample
complexity of dictionary learning for tensor data can be significantly lower
than that for unstructured data.
|
Option Discovery in Hierarchical Reinforcement Learning using
Spatio-Temporal Clustering | This paper introduces an automated skill acquisition framework in
reinforcement learning which involves identifying a hierarchical description of
the given task in terms of abstract states and extended actions between
abstract states. Identifying such structures present in the task provides ways
to simplify and speed up reinforcement learning algorithms. These structures
also help to generalize such algorithms over multiple tasks without relearning
policies from scratch. We use ideas from dynamical systems to find metastable
regions in the state space and associate them with abstract states. The
spectral clustering algorithm PCCA+ is used to identify suitable abstractions
aligned to the underlying structure. Skills are defined in terms of the
sequence of actions that lead to transitions between such abstract states. The
connectivity information from PCCA+ is used to generate these skills or
options. These skills are independent of the learning task and can be
efficiently reused across a variety of tasks defined over the same model. This
approach works well even without the exact model of the environment by using
sample trajectories to construct an approximate estimate. We also present our
approach to scaling the skill acquisition framework to complex tasks with large
state spaces for which we perform state aggregation using the representation
learned from an action conditional video prediction network and use the skill
acquisition framework on the aggregated state space.
|
Yelp Dataset Challenge: Review Rating Prediction | Review websites, such as TripAdvisor and Yelp, allow users to post online
reviews for various businesses, products and services, and have been recently
shown to have a significant influence on consumer shopping behaviour. An online
review typically consists of free-form text and a star rating out of 5. The
problem of predicting a user's star rating for a product, given the user's text
review for that product, is called Review Rating Prediction and has lately
become a popular, albeit hard, problem in machine learning. In this paper, we
treat Review Rating Prediction as a multi-class classification problem, and
build sixteen different prediction models by combining four feature extraction
methods, (i) unigrams, (ii) bigrams, (iii) trigrams and (iv) Latent Semantic
Indexing, with four machine learning algorithms, (i) logistic regression, (ii)
Naive Bayes classification, (iii) perceptrons, and (iv) linear Support Vector
Classification. We analyse the performance of each of these sixteen models to
come up with the best model for predicting the ratings from reviews. We use the
dataset provided by Yelp for training and testing the models.
|
Dynamic Frame skip Deep Q Network | Deep Reinforcement Learning methods have achieved state of the art
performance in learning control policies for the games in the Atari 2600
domain. One of the important parameters in the Arcade Learning Environment
(ALE) is the frame skip rate. It decides the granularity at which agents can
control game play. A frame skip value of $k$ allows the agent to repeat a
selected action $k$ number of times. The current state of the art architectures
like Deep Q-Network (DQN) and Dueling Network Architectures (DuDQN) consist of
a framework with a static frame skip rate, where the action output from the
network is repeated for a fixed number of frames regardless of the current
state. In this paper, we propose a new architecture, Dynamic Frame skip Deep
Q-Network (DFDQN) which makes the frame skip rate a dynamic learnable
parameter. This allows us to choose the number of times an action is to be
repeated based on the current state. We show empirically that such a setting
improves the performance on relatively harder games like Seaquest.
|
Deep Action Sequence Learning for Causal Shape Transformation | Deep learning became the method of choice in recent year for solving a wide
variety of predictive analytics tasks. For sequence prediction, recurrent
neural networks (RNN) are often the go-to architecture for exploiting
sequential information where the output is dependent on previous computation.
However, the dependencies of the computation lie in the latent domain which may
not be suitable for certain applications involving the prediction of a
step-wise transformation sequence that is dependent on the previous computation
only in the visible domain. We propose that a hybrid architecture of
convolution neural networks (CNN) and stacked autoencoders (SAE) is sufficient
to learn a sequence of actions that nonlinearly transforms an input shape or
distribution into a target shape or distribution with the same support. While
such a framework can be useful in a variety of problems such as robotic path
planning, sequential decision-making in games, and identifying material
processing pathways to achieve desired microstructures, the application of the
framework is exemplified by the control of fluid deformations in a microfluidic
channel by deliberately placing a sequence of pillars. Learning of a multistep
topological transform has significant implications for rapid advances in
material science and biomedical applications.
|
On the Evaluation of Dialogue Systems with Next Utterance Classification | An open challenge in constructing dialogue systems is developing methods for
automatically learning dialogue strategies from large amounts of unlabelled
data. Recent work has proposed Next-Utterance-Classification (NUC) as a
surrogate task for building dialogue systems from text data. In this paper we
investigate the performance of humans on this task to validate the relevance of
NUC as a method of evaluation. Our results show three main findings: (1) humans
are able to correctly classify responses at a rate much better than chance,
thus confirming that the task is feasible, (2) human performance levels vary
across task domains (we consider 3 datasets) and expertise levels (novice vs
experts), thus showing that a range of performance is possible on this type of
task, (3) automated dialogue systems built using state-of-the-art machine
learning methods have similar performance to the human novices, but worse than
the experts, thus confirming the utility of this class of tasks for driving
further research in automated dialogue systems.
|
Optimization Beyond Prediction: Prescriptive Price Optimization | This paper addresses a novel data science problem, prescriptive price
optimization, which derives the optimal price strategy to maximize future
profit/revenue on the basis of massive predictive formulas produced by machine
learning. The prescriptive price optimization first builds sales forecast
formulas of multiple products, on the basis of historical data, which reveal
complex relationships between sales and prices, such as price elasticity of
demand and cannibalization. Then, it constructs a mathematical optimization
problem on the basis of those predictive formulas. We present that the
optimization problem can be formulated as an instance of binary quadratic
programming (BQP). Although BQP problems are NP-hard in general and
computationally intractable, we propose a fast approximation algorithm using a
semi-definite programming (SDP) relaxation, which is closely related to the
Goemans-Williamson's Max-Cut approximation. Our experiments on simulation and
real retail datasets show that our prescriptive price optimization
simultaneously derives the optimal prices of tens/hundreds products with
practical computational time, that potentially improve 8.2% of gross profit of
those products.
|
Learning activation functions from data using cubic spline interpolation | Neural networks require a careful design in order to perform properly on a
given task. In particular, selecting a good activation function (possibly in a
data-dependent fashion) is a crucial step, which remains an open problem in the
research community. Despite a large amount of investigations, most current
implementations simply select one fixed function from a small set of
candidates, which is not adapted during training, and is shared among all
neurons throughout the different layers. However, neither two of these
assumptions can be supposed optimal in practice. In this paper, we present a
principled way to have data-dependent adaptation of the activation functions,
which is performed independently for each neuron. This is achieved by
leveraging over past and present advances on cubic spline interpolation,
allowing for local adaptation of the functions around their regions of use. The
resulting algorithm is relatively cheap to implement, and overfitting is
counterbalanced by the inclusion of a novel damping criterion, which penalizes
unwanted oscillations from a predefined shape. Experimental results validate
the proposal over two well-known benchmarks.
|
Detecting Novel Processes with CANDIES -- An Holistic Novelty Detection
Technique based on Probabilistic Models | In this article, we propose CANDIES (Combined Approach for Novelty Detection
in Intelligent Embedded Systems), a new approach to novelty detection in
technical systems. We assume that in a technical system several processes
interact. If we observe these processes with sensors, we are able to model the
observations (samples) with a probabilistic model, where, in an ideal case, the
components of the parametric mixture density model we use, correspond to the
processes in the real world. Eventually, at run-time, novel processes emerge in
the technical systems such as in the case of an unpredictable failure. As a
consequence, new kinds of samples are observed that require an adaptation of
the model. CANDIES relies on mixtures of Gaussians which can be used for
classification purposes, too. New processes may emerge in regions of the
models' input spaces where few samples were observed before (low-density
regions) or in regions where already many samples were available (high-density
regions). The latter case is more difficult, but most existing solutions focus
on the former. Novelty detection in low- and high-density regions requires
different detection strategies. With CANDIES, we introduce a new technique to
detect novel processes in high-density regions by means of a fast online
goodness-of-fit test. For detection in low-density regions we combine this
approach with a 2SND (Two-Stage-Novelty-Detector) which we presented in
preliminary work. The properties of CANDIES are evaluated using artificial data
and benchmark data from the field of intrusion detection in computer networks,
where the task is to detect new kinds of attacks.
|
Active Learning On Weighted Graphs Using Adaptive And Non-adaptive
Approaches | This paper studies graph-based active learning, where the goal is to
reconstruct a binary signal defined on the nodes of a weighted graph, by
sampling it on a small subset of the nodes. A new sampling algorithm is
proposed, which sequentially selects the graph nodes to be sampled, based on an
aggressive search for the boundary of the signal over the graph. The algorithm
generalizes a recent method for sampling nodes in unweighted graphs. The
generalization improves the sampling performance using the information gained
from the available graph weights. An analysis of the number of samples required
by the proposed algorithm is provided, and the gain over the unweighted method
is further demonstrated in simulations. Additionally, the proposed method is
compared with an alternative state of-the-art method, which is based on the
graph's spectral properties. It is shown that the proposed method significantly
outperforms the spectral sampling method, if the signal needs to be predicted
with high accuracy. On the other hand, if a higher level of inaccuracy is
tolerable, then the spectral method outperforms the proposed aggressive search
method. Consequently, we propose a hybrid method, which is shown to combine the
advantages of both approaches.
|
Linearized GMM Kernels and Normalized Random Fourier Features | The method of "random Fourier features (RFF)" has become a popular tool for
approximating the "radial basis function (RBF)" kernel. The variance of RFF is
actually large. Interestingly, the variance can be substantially reduced by a
simple normalization step as we theoretically demonstrate. We name the improved
scheme as the "normalized RFF (NRFF)".
We also propose the "generalized min-max (GMM)" kernel as a measure of data
similarity. GMM is positive definite as there is an associated hashing method
named "generalized consistent weighted sampling (GCWS)" which linearizes this
nonlinear kernel. We provide an extensive empirical evaluation of the RBF
kernel and the GMM kernel on more than 50 publicly available datasets. For a
majority of the datasets, the (tuning-free) GMM kernel outperforms the
best-tuned RBF kernel.
We conduct extensive experiments for comparing the linearized RBF kernel
using NRFF with the linearized GMM kernel using GCWS. We observe that, to reach
a comparable classification accuracy, GCWS typically requires substantially
fewer samples than NRFF, even on datasets where the original RBF kernel
outperforms the original GMM kernel. The empirical success of GCWS (compared to
NRFF) can also be explained from a theoretical perspective. Firstly, the
relative variance (normalized by the squared expectation) of GCWS is
substantially smaller than that of NRFF, except for the very high similarity
region (where the variances of both methods are close to zero). Secondly, if we
make a model assumption on the data, we can show analytically that GCWS
exhibits much smaller variance than NRFF for estimating the same object (e.g.,
the RBF kernel), except for the very high similarity region.
|
Supervised Learning with Quantum-Inspired Tensor Networks | Tensor networks are efficient representations of high-dimensional tensors
which have been very successful for physics and mathematics applications. We
demonstrate how algorithms for optimizing such networks can be adapted to
supervised learning tasks by using matrix product states (tensor trains) to
parameterize models for classifying images. For the MNIST data set we obtain
less than 1% test set classification error. We discuss how the tensor network
form imparts additional structure to the learned model and suggest a possible
generative interpretation.
|
Recurrent Exponential-Family Harmoniums without Backprop-Through-Time | Exponential-family harmoniums (EFHs), which extend restricted Boltzmann
machines (RBMs) from Bernoulli random variables to other exponential families
(Welling et al., 2005), are generative models that can be trained with
unsupervised-learning techniques, like contrastive divergence (Hinton et al.
2006; Hinton, 2002), as density estimators for static data. Methods for
extending RBMs--and likewise EFHs--to data with temporal dependencies have been
proposed previously (Sutskever and Hinton, 2007; Sutskever et al., 2009), the
learning procedure being validated by qualitative assessment of the generative
model. Here we propose and justify, from a very different perspective, an
alternative training procedure, proving sufficient conditions for optimal
inference under that procedure. The resulting algorithm can be learned with
only forward passes through the data--backprop-through-time is not required, as
in previous approaches. The proof exploits a recent result about information
retention in density estimators (Makin and Sabes, 2015), and applies it to a
"recurrent EFH" (rEFH) by induction. Finally, we demonstrate optimality by
simulation, testing the rEFH: (1) as a filter on training data generated with a
linear dynamical system, the position of which is noisily reported by a
population of "neurons" with Poisson-distributed spike counts; and (2) with the
qualitative experiments proposed by Sutskever et al. (2009).
|
Declarative Machine Learning - A Classification of Basic Properties and
Types | Declarative machine learning (ML) aims at the high-level specification of ML
tasks or algorithms, and automatic generation of optimized execution plans from
these specifications. The fundamental goal is to simplify the usage and/or
development of ML algorithms, which is especially important in the context of
large-scale computations. However, ML systems at different abstraction levels
have emerged over time and accordingly there has been a controversy about the
meaning of this general definition of declarative ML. Specification
alternatives range from ML algorithms expressed in domain-specific languages
(DSLs) with optimization for performance, to ML task (learning problem)
specifications with optimization for performance and accuracy. We argue that
these different types of declarative ML complement each other as they address
different users (data scientists and end users). This paper makes an attempt to
create a taxonomy for declarative ML, including a definition of essential basic
properties and types of declarative ML. Along the way, we provide insights into
implications of these properties. We also use this taxonomy to classify
existing systems. Finally, we draw conclusions on defining appropriate
benchmarks and specification languages for declarative ML.
|
A Multi-Batch L-BFGS Method for Machine Learning | The question of how to parallelize the stochastic gradient descent (SGD)
method has received much attention in the literature. In this paper, we focus
instead on batch methods that use a sizeable fraction of the training set at
each iteration to facilitate parallelism, and that employ second-order
information. In order to improve the learning process, we follow a multi-batch
approach in which the batch changes at each iteration. This can cause
difficulties because L-BFGS employs gradient differences to update the Hessian
approximations, and when these gradients are computed using different data
points the process can be unstable. This paper shows how to perform stable
quasi-Newton updating in the multi-batch setting, illustrates the behavior of
the algorithm in a distributed computing platform, and studies its convergence
properties for both the convex and nonconvex cases.
|
One-shot Learning with Memory-Augmented Neural Networks | Despite recent breakthroughs in the applications of deep neural networks, one
setting that presents a persistent challenge is that of "one-shot learning."
Traditional gradient-based networks require a lot of data to learn, often
through extensive iterative training. When new data is encountered, the models
must inefficiently relearn their parameters to adequately incorporate the new
information without catastrophic interference. Architectures with augmented
memory capacities, such as Neural Turing Machines (NTMs), offer the ability to
quickly encode and retrieve new information, and hence can potentially obviate
the downsides of conventional models. Here, we demonstrate the ability of a
memory-augmented neural network to rapidly assimilate new data, and leverage
this data to make accurate predictions after only a few samples. We also
introduce a new method for accessing an external memory that focuses on memory
content, unlike previous methods that additionally use memory location-based
focusing mechanisms.
|
A Hierarchical Latent Variable Encoder-Decoder Model for Generating
Dialogues | Sequential data often possesses a hierarchical structure with complex
dependencies between subsequences, such as found between the utterances in a
dialogue. In an effort to model this kind of generative process, we propose a
neural network-based generative architecture, with latent stochastic variables
that span a variable number of time steps. We apply the proposed model to the
task of dialogue response generation and compare it with recent neural network
architectures. We evaluate the model performance through automatic evaluation
metrics and by carrying out a human evaluation. The experiments demonstrate
that our model improves upon recently proposed models and that the latent
variables facilitate the generation of long outputs and maintain the context.
|
On a convergent off -policy temporal difference learning algorithm in
on-line learning environment | In this paper we provide a rigorous convergence analysis of a "off"-policy
temporal difference learning algorithm with linear function approximation and
per time-step linear computational complexity in "online" learning environment.
The algorithm considered here is TDC with importance weighting introduced by
Maei et al. We support our theoretical results by providing suitable empirical
results for standard off-policy counterexamples.
|
Inter-Battery Topic Representation Learning | In this paper, we present the Inter-Battery Topic Model (IBTM). Our approach
extends traditional topic models by learning a factorized latent variable
representation. The structured representation leads to a model that marries
benefits traditionally associated with a discriminative approach, such as
feature selection, with those of a generative model, such as principled
regularization and ability to handle missing data. The factorization is
provided by representing data in terms of aligned pairs of observations as
different views. This provides means for selecting a representation that
separately models topics that exist in both views from the topics that are
unique to a single view. This structured consolidation allows for efficient and
robust inference and provides a compact and efficient representation. Learning
is performed in a Bayesian fashion by maximizing a rigorous bound on the
log-likelihood. Firstly, we illustrate the benefits of the model on a synthetic
dataset,. The model is then evaluated in both uni- and multi-modality settings
on two different classification tasks with off-the-shelf convolutional neural
network (CNN) features which generate state-of-the-art results with extremely
compact representations.
|
Evaluation System for a Bayesian Optimization Service | Bayesian optimization is an elegant solution to the hyperparameter
optimization problem in machine learning. Building a reliable and robust
Bayesian optimization service requires careful testing methodology and sound
statistical analysis. In this talk we will outline our development of an
evaluation framework to rigorously test and measure the impact of changes to
the SigOpt optimization service. We present an overview of our evaluation
system and discuss how this framework empowers our research engineers to
confidently and quickly make changes to our core optimization engine
|
Variational hybridization and transformation for large inaccurate
noisy-or networks | Variational inference provides approximations to the computationally
intractable posterior distribution in Bayesian networks. A prominent medical
application of noisy-or Bayesian network is to infer potential diseases given
observed symptoms. Previous studies focus on approximating a handful of
complicated pathological cases using variational transformation. Our goal is to
use variational transformation as part of a novel hybridized inference for
serving reliable and real time diagnosis at web scale. We propose a hybridized
inference that allows variational parameters to be estimated without disease
posteriors or priors, making the inference faster and much of its computation
recyclable. In addition, we propose a transformation ranking algorithm that is
very stable to large variances in network prior probabilities, a common issue
that arises in medical applications of Bayesian networks. In experiments, we
perform comparative study on a large real life medical network and scalability
study on a much larger (36,000x) synthesized network.
|
Adversarial Delays in Online Strongly-Convex Optimization | We consider the problem of strongly-convex online optimization in presence of
adversarial delays; in a T-iteration online game, the feedback of the player's
query at time t is arbitrarily delayed by an adversary for d_t rounds and
delivered before the game ends, at iteration t+d_t-1. Specifically for
\algo{online-gradient-descent} algorithm we show it has a simple regret bound
of \Oh{\sum_{t=1}^T \log (1+ \frac{d_t}{t})}. This gives a clear and simple
bound without resorting any distributional and limiting assumptions on the
delays. We further show how this result encompasses and generalizes several of
the existing known results in the literature. Specifically it matches the
celebrated logarithmic regret \Oh{\log T} when there are no delays (i.e. d_t =
1) and regret bound of \Oh{\tau \log T} for constant delays d_t = \tau.
|
Faster Projection-free Convex Optimization over the Spectrahedron | Minimizing a convex function over the spectrahedron, i.e., the set of all
positive semidefinite matrices with unit trace, is an important optimization
task with many applications in optimization, machine learning, and signal
processing. It is also notoriously difficult to solve in large-scale since
standard techniques require expensive matrix decompositions. An alternative, is
the conditional gradient method (aka Frank-Wolfe algorithm) that regained much
interest in recent years, mostly due to its application to this specific
setting. The key benefit of the CG method is that it avoids expensive matrix
decompositions all together, and simply requires a single eigenvector
computation per iteration, which is much more efficient. On the downside, the
CG method, in general, converges with an inferior rate. The error for
minimizing a $\beta$-smooth function after $t$ iterations scales like
$\beta/t$. This convergence rate does not improve even if the function is also
strongly convex.
In this work we present a modification of the CG method tailored for convex
optimization over the spectrahedron. The per-iteration complexity of the method
is essentially identical to that of the standard CG method: only a single
eigenvecor computation is required. For minimizing an $\alpha$-strongly convex
and $\beta$-smooth function, the expected approximation error of the method
after $t$ iterations is: $$O\left({\min\{\frac{\beta{}}{t}
,\left({\frac{\beta\sqrt{\textrm{rank}(\textbf{X}^*)}}{\alpha^{1/4}t}}\right)^{4/3},
\left({\frac{\beta}{\sqrt{\alpha}\lambda_{\min}(\textbf{X}^*)t}}\right)^{2}\}}\right)
,$$ where $\textbf{X}^*$ is the optimal solution. To the best of our knowledge,
this is the first result that attains provably faster convergence rates for a
CG variant for optimization over the spectrahedron. We also present encouraging
preliminary empirical results.
|
Convergence of Contrastive Divergence with Annealed Learning Rate in
Exponential Family | In our recent paper, we showed that in exponential family, contrastive
divergence (CD) with fixed learning rate will give asymptotically consistent
estimates \cite{wu2016convergence}. In this paper, we establish consistency and
convergence rate of CD with annealed learning rate $\eta_t$. Specifically,
suppose CD-$m$ generates the sequence of parameters $\{\theta_t\}_{t \ge 0}$
using an i.i.d. data sample $\mathbf{X}_1^n \sim p_{\theta^*}$ of size $n$,
then $\delta_n(\mathbf{X}_1^n) = \limsup_{t \to \infty} \Vert \sum_{s=t_0}^t
\eta_s \theta_s / \sum_{s=t_0}^t \eta_s - \theta^* \Vert$ converges in
probability to 0 at a rate of $1/\sqrt[3]{n}$. The number ($m$) of MCMC
transitions in CD only affects the coefficient factor of convergence rate. Our
proof is not a simple extension of the one in \cite{wu2016convergence}. which
depends critically on the fact that $\{\theta_t\}_{t \ge 0}$ is a homogeneous
Markov chain conditional on the observed sample $\mathbf{X}_1^n$. Under
annealed learning rate, the homogeneous Markov property is not available and we
have to develop an alternative approach based on super-martingales. Experiment
results of CD on a fully-visible $2\times 2$ Boltzmann Machine are provided to
demonstrate our theoretical results.
|
End-to-End Kernel Learning with Supervised Convolutional Kernel Networks | In this paper, we introduce a new image representation based on a multilayer
kernel machine. Unlike traditional kernel methods where data representation is
decoupled from the prediction task, we learn how to shape the kernel with
supervision. We proceed by first proposing improvements of the
recently-introduced convolutional kernel networks (CKNs) in the context of
unsupervised learning; then, we derive backpropagation rules to take advantage
of labeled training data. The resulting model is a new type of convolutional
neural network, where optimizing the filters at each layer is equivalent to
learning a linear subspace in a reproducing kernel Hilbert space (RKHS). We
show that our method achieves reasonably competitive performance for image
classification on some standard "deep learning" datasets such as CIFAR-10 and
SVHN, and also for image super-resolution, demonstrating the applicability of
our approach to a large variety of image-related tasks.
|
Piece-wise quadratic approximations of arbitrary error functions for
fast and robust machine learning | Most of machine learning approaches have stemmed from the application of
minimizing the mean squared distance principle, based on the computationally
efficient quadratic optimization methods. However, when faced with
high-dimensional and noisy data, the quadratic error functionals demonstrated
many weaknesses including high sensitivity to contaminating factors and
dimensionality curse. Therefore, a lot of recent applications in machine
learning exploited properties of non-quadratic error functionals based on $L_1$
norm or even sub-linear potentials corresponding to quasinorms $L_p$ ($0<p<1$).
The back side of these approaches is increase in computational cost for
optimization. Till so far, no approaches have been suggested to deal with {\it
arbitrary} error functionals, in a flexible and computationally efficient
framework. In this paper, we develop a theory and basic universal data
approximation algorithms ($k$-means, principal components, principal manifolds
and graphs, regularized and sparse regression), based on piece-wise quadratic
error potentials of subquadratic growth (PQSQ potentials). We develop a new and
universal framework to minimize {\it arbitrary sub-quadratic error potentials}
using an algorithm with guaranteed fast convergence to the local or global
error minimum. The theory of PQSQ potentials is based on the notion of the cone
of minorant functions, and represents a natural approximation formalism based
on the application of min-plus algebra. The approach can be applied in most of
existing machine learning methods, including methods of data approximation and
regularized and sparse regression, leading to the improvement in the
computational cost/accuracy trade-off. We demonstrate that on synthetic and
real-life datasets PQSQ-based machine learning methods achieve orders of
magnitude faster computational performance than the corresponding
state-of-the-art methods.
|
On the Robustness of Decision Tree Learning under Label Noise | In most practical problems of classifier learning, the training data suffers
from the label noise. Hence, it is important to understand how robust is a
learning algorithm to such label noise. This paper presents some theoretical
analysis to show that many popular decision tree algorithms are robust to
symmetric label noise under large sample size. We also present some sample
complexity results which provide some bounds on the sample size for the
robustness to hold with a high probability. Through extensive simulations we
illustrate this robustness.
|
Unsupervised Feature Extraction by Time-Contrastive Learning and
Nonlinear ICA | Nonlinear independent component analysis (ICA) provides an appealing
framework for unsupervised feature learning, but the models proposed so far are
not identifiable. Here, we first propose a new intuitive principle of
unsupervised deep learning from time series which uses the nonstationary
structure of the data. Our learning principle, time-contrastive learning (TCL),
finds a representation which allows optimal discrimination of time segments
(windows). Surprisingly, we show how TCL can be related to a nonlinear ICA
model, when ICA is redefined to include temporal nonstationarities. In
particular, we show that TCL combined with linear ICA estimates the nonlinear
ICA model up to point-wise transformations of the sources, and this solution is
unique --- thus providing the first identifiability result for nonlinear ICA
which is rigorous, constructive, as well as very general.
|
Fast $\epsilon$-free Inference of Simulation Models with Bayesian
Conditional Density Estimation | Many statistical models can be simulated forwards but have intractable
likelihoods. Approximate Bayesian Computation (ABC) methods are used to infer
properties of these models from data. Traditionally these methods approximate
the posterior over parameters by conditioning on data being inside an
$\epsilon$-ball around the observed data, which is only correct in the limit
$\epsilon\!\rightarrow\!0$. Monte Carlo methods can then draw samples from the
approximate posterior to approximate predictions or error bars on parameters.
These algorithms critically slow down as $\epsilon\!\rightarrow\!0$, and in
practice draw samples from a broader distribution than the posterior. We
propose a new approach to likelihood-free inference based on Bayesian
conditional density estimation. Preliminary inferences based on limited
simulation data are used to guide later simulations. In some cases, learning an
accurate parametric representation of the entire true posterior distribution
requires fewer model simulations than Monte Carlo ABC methods need to produce a
single sample from an approximate posterior.
|
Towards Automation of Knowledge Understanding: An Approach for
Probabilistic Generative Classifiers | After data selection, pre-processing, transformation, and feature extraction,
knowledge extraction is not the final step in a data mining process. It is then
necessary to understand this knowledge in order to apply it efficiently and
effectively. Up to now, there is a lack of appropriate techniques that support
this significant step. This is partly due to the fact that the assessment of
knowledge is often highly subjective, e.g., regarding aspects such as novelty
or usefulness. These aspects depend on the specific knowledge and requirements
of the data miner. There are, however, a number of aspects that are objective
and for which it is possible to provide appropriate measures. In this article
we focus on classification problems and use probabilistic generative
classifiers based on mixture density models that are quite common in data
mining applications. We define objective measures to assess the
informativeness, uniqueness, importance, discrimination, representativity,
uncertainty, and distinguishability of rules contained in these classifiers
numerically. These measures not only support a data miner in evaluating results
of a data mining process based on such classifiers. As we will see in
illustrative case studies, they may also be used to improve the data mining
process itself or to support the later application of the extracted knowledge.
|
Deep Multi-task Representation Learning: A Tensor Factorisation Approach | Most contemporary multi-task learning methods assume linear models. This
setting is considered shallow in the era of deep learning. In this paper, we
present a new deep multi-task representation learning framework that learns
cross-task sharing structure at every layer in a deep network. Our approach is
based on generalising the matrix factorisation techniques explicitly or
implicitly used by many conventional MTL algorithms to tensor factorisation, to
realise automatic learning of end-to-end knowledge sharing in deep networks.
This is in contrast to existing deep learning approaches that need a
user-defined multi-task sharing strategy. Our approach applies to both
homogeneous and heterogeneous MTL. Experiments demonstrate the efficacy of our
deep multi-task representation learning in terms of both higher accuracy and
fewer design choices.
|
Bayesian Hyperparameter Optimization for Ensemble Learning | In this paper, we bridge the gap between hyperparameter optimization and
ensemble learning by performing Bayesian optimization of an ensemble with
regards to its hyperparameters. Our method consists in building a fixed-size
ensemble, optimizing the configuration of one classifier of the ensemble at
each iteration of the hyperparameter optimization algorithm, taking into
consideration the interaction with the other models when evaluating potential
performances. We also consider the case where the ensemble is to be
reconstructed at the end of the hyperparameter optimization phase, through a
greedy selection over the pool of models generated during the optimization. We
study the performance of our proposed method on three different hyperparameter
spaces, showing that our approach is better than both the best single model and
a greedy ensemble construction over the models produced by a standard Bayesian
optimization.
|
Stochastic Variance Reduction Methods for Saddle-Point Problems | We consider convex-concave saddle-point problems where the objective
functions may be split in many components, and extend recent stochastic
variance reduction methods (such as SVRG or SAGA) to provide the first
large-scale linearly convergent algorithms for this class of problems which is
common in machine learning. While the algorithmic extension is straightforward,
it comes with challenges and opportunities: (a) the convex minimization
analysis does not apply and we use the notion of monotone operators to prove
convergence, showing in particular that the same algorithm applies to a larger
class of problems, such as variational inequalities, (b) there are two notions
of splits, in terms of functions, or in terms of partial derivatives, (c) the
split does need to be done with convex-concave terms, (d) non-uniform sampling
is key to an efficient algorithm, both in theory and practice, and (e) these
incremental algorithms can be easily accelerated using a simple extension of
the "catalyst" framework, leading to an algorithm which is always superior to
accelerated batch algorithms.
|
Ristretto: Hardware-Oriented Approximation of Convolutional Neural
Networks | Convolutional neural networks (CNN) have achieved major breakthroughs in
recent years. Their performance in computer vision have matched and in some
areas even surpassed human capabilities. Deep neural networks can capture
complex non-linear features; however this ability comes at the cost of high
computational and memory requirements. State-of-art networks require billions
of arithmetic operations and millions of parameters. To enable embedded devices
such as smartphones, Google glasses and monitoring cameras with the astonishing
power of deep learning, dedicated hardware accelerators can be used to decrease
both execution time and power consumption. In applications where fast
connection to the cloud is not guaranteed or where privacy is important,
computation needs to be done locally. Many hardware accelerators for deep
neural networks have been proposed recently. A first important step of
accelerator design is hardware-oriented approximation of deep networks, which
enables energy-efficient inference. We present Ristretto, a fast and automated
framework for CNN approximation. Ristretto simulates the hardware arithmetic of
a custom hardware accelerator. The framework reduces the bit-width of network
parameters and outputs of resource-intense layers, which reduces the chip area
for multiplication units significantly. Alternatively, Ristretto can remove the
need for multipliers altogether, resulting in an adder-only arithmetic. The
tool fine-tunes trimmed networks to achieve high classification accuracy. Since
training of deep neural networks can be time-consuming, Ristretto uses highly
optimized routines which run on the GPU. This enables fast compression of any
given network. Given a maximum tolerance of 1%, Ristretto can successfully
condense CaffeNet and SqueezeNet to 8-bit. The code for Ristretto is available.
|
Data-driven root-cause analysis for distributed system anomalies | Modern distributed cyber-physical systems encounter a large variety of
anomalies and in many cases, they are vulnerable to catastrophic fault
propagation scenarios due to strong connectivity among the sub-systems. In this
regard, root-cause analysis becomes highly intractable due to complex fault
propagation mechanisms in combination with diverse operating modes. This paper
presents a new data-driven framework for root-cause analysis for addressing
such issues. The framework is based on a spatiotemporal feature extraction
scheme for distributed cyber-physical systems built on the concept of symbolic
dynamics for discovering and representing causal interactions among subsystems
of a complex system. We present two approaches for root-cause analysis, namely
the sequential state switching ($S^3$, based on free energy concept of a
Restricted Boltzmann Machine, RBM) and artificial anomaly association ($A^3$, a
multi-class classification framework using deep neural networks, DNN).
Synthetic data from cases with failed pattern(s) and anomalous node are
simulated to validate the proposed approaches, then compared with the
performance of vector autoregressive (VAR) model-based root-cause analysis.
Real dataset based on Tennessee Eastman process (TEP) is also used for
validation. The results show that: (1) $S^3$ and $A^3$ approaches can obtain
high accuracy in root-cause analysis and successfully handle multiple nominal
operation modes, and (2) the proposed tool-chain is shown to be scalable while
maintaining high accuracy.
|
Fast Randomized Semi-Supervised Clustering | We consider the problem of clustering partially labeled data from a minimal
number of randomly chosen pairwise comparisons between the items. We introduce
an efficient local algorithm based on a power iteration of the non-backtracking
operator and study its performance on a simple model. For the case of two
clusters, we give bounds on the classification error and show that a small
error can be achieved from $O(n)$ randomly chosen measurements, where $n$ is
the number of items in the dataset. Our algorithm is therefore efficient both
in terms of time and space complexities. We also investigate numerically the
performance of the algorithm on synthetic and real world data.
|
Residual Networks Behave Like Ensembles of Relatively Shallow Networks | In this work we propose a novel interpretation of residual networks showing
that they can be seen as a collection of many paths of differing length.
Moreover, residual networks seem to enable very deep networks by leveraging
only the short paths during training. To support this observation, we rewrite
residual networks as an explicit collection of paths. Unlike traditional
models, paths through residual networks vary in length. Further, a lesion study
reveals that these paths show ensemble-like behavior in the sense that they do
not strongly depend on each other. Finally, and most surprising, most paths are
shorter than one might expect, and only the short paths are needed during
training, as longer paths do not contribute any gradient. For example, most of
the gradient in a residual network with 110 layers comes from paths that are
only 10-34 layers deep. Our results reveal one of the key characteristics that
seem to enable the training of very deep networks: Residual networks avoid the
vanishing gradient problem by introducing short paths which can carry gradient
throughout the extent of very deep networks.
|
Deep Variational Bayes Filters: Unsupervised Learning of State Space
Models from Raw Data | We introduce Deep Variational Bayes Filters (DVBF), a new method for
unsupervised learning and identification of latent Markovian state space
models. Leveraging recent advances in Stochastic Gradient Variational Bayes,
DVBF can overcome intractable inference distributions via variational
inference. Thus, it can handle highly nonlinear input data with temporal and
spatial dependencies such as image sequences without domain knowledge. Our
experiments show that enabling backpropagation through transitions enforces
state space assumptions and significantly improves information content of the
latent embedding. This also enables realistic long-term prediction.
|
Combining Adversarial Guarantees and Stochastic Fast Rates in Online
Learning | We consider online learning algorithms that guarantee worst-case regret rates
in adversarial environments (so they can be deployed safely and will perform
robustly), yet adapt optimally to favorable stochastic environments (so they
will perform well in a variety of settings of practical importance). We
quantify the friendliness of stochastic environments by means of the well-known
Bernstein (a.k.a. generalized Tsybakov margin) condition. For two recent
algorithms (Squint for the Hedge setting and MetaGrad for online convex
optimization) we show that the particular form of their data-dependent
individual-sequence regret guarantees implies that they adapt automatically to
the Bernstein parameters of the stochastic environment. We prove that these
algorithms attain fast rates in their respective settings both in expectation
and with high probability.
|
Structured Prediction Theory Based on Factor Graph Complexity | We present a general theoretical analysis of structured prediction with a
series of new results. We give new data-dependent margin guarantees for
structured prediction for a very wide family of loss functions and a general
family of hypotheses, with an arbitrary factor graph decomposition. These are
the tightest margin bounds known for both standard multi-class and general
structured prediction problems. Our guarantees are expressed in terms of a
data-dependent complexity measure, factor graph complexity, which we show can
be estimated from data and bounded in terms of familiar quantities. We further
extend our theory by leveraging the principle of Voted Risk Minimization (VRM)
and show that learning is possible even with complex factor graphs. We present
new learning bounds for this advanced setting, which we use to design two new
algorithms, Voted Conditional Random Field (VCRF) and Voted Structured Boosting
(StructBoost). These algorithms can make use of complex features and factor
graphs and yet benefit from favorable learning guarantees. We also report the
results of experiments with VCRF on several datasets to validate our theory.
|
Unreasonable Effectiveness of Learning Neural Networks: From Accessible
States and Robust Ensembles to Basic Algorithmic Schemes | In artificial neural networks, learning from data is a computationally
demanding task in which a large number of connection weights are iteratively
tuned through stochastic-gradient-based heuristic processes over a
cost-function. It is not well understood how learning occurs in these systems,
in particular how they avoid getting trapped in configurations with poor
computational performance. Here we study the difficult case of networks with
discrete weights, where the optimization landscape is very rough even for
simple architectures, and provide theoretical and numerical evidence of the
existence of rare - but extremely dense and accessible - regions of
configurations in the network weight space. We define a novel measure, which we
call the "robust ensemble" (RE), which suppresses trapping by isolated
configurations and amplifies the role of these dense regions. We analytically
compute the RE in some exactly solvable models, and also provide a general
algorithmic scheme which is straightforward to implement: define a
cost-function given by a sum of a finite number of replicas of the original
cost-function, with a constraint centering the replicas around a driving
assignment. To illustrate this, we derive several powerful new algorithms,
ranging from Markov Chains to message passing to gradient descent processes,
where the algorithms target the robust dense states, resulting in substantial
improvements in performance. The weak dependence on the number of precision
bits of the weights leads us to conjecture that very similar reasoning applies
to more conventional neural networks. Analogous algorithmic schemes can also be
applied to other optimization problems.
|
Query-Efficient Imitation Learning for End-to-End Autonomous Driving | One way to approach end-to-end autonomous driving is to learn a policy
function that maps from a sensory input, such as an image frame from a
front-facing camera, to a driving action, by imitating an expert driver, or a
reference policy. This can be done by supervised learning, where a policy
function is tuned to minimize the difference between the predicted and
ground-truth actions. A policy function trained in this way however is known to
suffer from unexpected behaviours due to the mismatch between the states
reachable by the reference policy and trained policy functions. More advanced
algorithms for imitation learning, such as DAgger, addresses this issue by
iteratively collecting training examples from both reference and trained
policies. These algorithms often requires a large number of queries to a
reference policy, which is undesirable as the reference policy is often
expensive. In this paper, we propose an extension of the DAgger, called
SafeDAgger, that is query-efficient and more suitable for end-to-end autonomous
driving. We evaluate the proposed SafeDAgger in a car racing simulator and show
that it indeed requires less queries to a reference policy. We observe a
significant speed up in convergence, which we conjecture to be due to the
effect of automated curriculum learning.
|
Virtual Worlds as Proxy for Multi-Object Tracking Analysis | Modern computer vision algorithms typically require expensive data
acquisition and accurate manual labeling. In this work, we instead leverage the
recent progress in computer graphics to generate fully labeled, dynamic, and
photo-realistic proxy virtual worlds. We propose an efficient real-to-virtual
world cloning method, and validate our approach by building and publicly
releasing a new video dataset, called Virtual KITTI (see
http://www.xrce.xerox.com/Research-Development/Computer-Vision/Proxy-Virtual-Worlds),
automatically labeled with accurate ground truth for object detection,
tracking, scene and instance segmentation, depth, and optical flow. We provide
quantitative experimental evidence suggesting that (i) modern deep learning
algorithms pre-trained on real data behave similarly in real and virtual
worlds, and (ii) pre-training on virtual data improves performance. As the gap
between real and virtual worlds is small, virtual worlds enable measuring the
impact of various weather and imaging conditions on recognition performance,
all other things being equal. We show these factors may affect drastically
otherwise high-performing deep models for tracking.
|
Swapout: Learning an ensemble of deep architectures | We describe Swapout, a new stochastic training method, that outperforms
ResNets of identical network structure yielding impressive results on CIFAR-10
and CIFAR-100. Swapout samples from a rich set of architectures including
dropout, stochastic depth and residual architectures as special cases. When
viewed as a regularization method swapout not only inhibits co-adaptation of
units in a layer, similar to dropout, but also across network layers. We
conjecture that swapout achieves strong regularization by implicitly tying the
parameters across layers. When viewed as an ensemble training method, it
samples a much richer set of architectures than existing methods such as
dropout or stochastic depth. We propose a parameterization that reveals
connections to exiting architectures and suggests a much richer set of
architectures to be explored. We show that our formulation suggests an
efficient training method and validate our conclusions on CIFAR-10 and
CIFAR-100 matching state of the art accuracy. Remarkably, our 32 layer wider
model performs similar to a 1001 layer ResNet model.
|
Regression with n$\to$1 by Expert Knowledge Elicitation | We consider regression under the "extremely small $n$ large $p$" condition,
where the number of samples $n$ is so small compared to the dimensionality $p$
that predictors cannot be estimated without prior knowledge. This setup occurs
in personalized medicine, for instance, when predicting treatment outcomes for
an individual patient based on noisy high-dimensional genomics data. A
remaining source of information is expert knowledge, which has received
relatively little attention in recent years. We formulate the inference problem
of asking expert feedback on features on a budget, propose an elicitation
strategy for a simple "small $n$" setting, and derive conditions under which
the elicitation strategy is optimal. Experiments on simulated experts, both on
synthetic and genomics data, demonstrate that the proposed strategy can
drastically improve prediction accuracy.
|
Deep Roots: Improving CNN Efficiency with Hierarchical Filter Groups | We propose a new method for creating computationally efficient and compact
convolutional neural networks (CNNs) using a novel sparse connection structure
that resembles a tree root. This allows a significant reduction in
computational cost and number of parameters compared to state-of-the-art deep
CNNs, without compromising accuracy, by exploiting the sparsity of inter-layer
filter dependencies. We validate our approach by using it to train more
efficient variants of state-of-the-art CNN architectures, evaluated on the
CIFAR10 and ILSVRC datasets. Our results show similar or higher accuracy than
the baseline architectures with much less computation, as measured by CPU and
GPU timings. For example, for ResNet 50, our model has 40% fewer parameters,
45% fewer floating point operations, and is 31% (12%) faster on a CPU (GPU).
For the deeper ResNet 200 our model has 25% fewer floating point operations and
44% fewer parameters, while maintaining state-of-the-art accuracy. For
GoogLeNet, our model has 7% fewer parameters and is 21% (16%) faster on a CPU
(GPU).
|
Linear-memory and Decomposition-invariant Linearly Convergent
Conditional Gradient Algorithm for Structured Polytopes | Recently, several works have shown that natural modifications of the
classical conditional gradient method (aka Frank-Wolfe algorithm) for
constrained convex optimization, provably converge with a linear rate when: i)
the feasible set is a polytope, and ii) the objective is smooth and
strongly-convex. However, all of these results suffer from two significant
shortcomings: large memory requirement due to the need to store an explicit
convex decomposition of the current iterate, and as a consequence, large
running-time overhead per iteration, and worst case convergence rate that
depends unfavorably on the dimension.
In this work we present a new conditional gradient variant and a
corresponding analysis that improves on both of the above shortcomings. In
particular: both memory and computation overheads are only linear in the
dimension. Moreover, in case the optimal solution is sparse, the new
convergence rate replaces a factor which is at least linear in the dimension in
previous works, with a linear dependence on the number of non-zeros in the
optimal solution.
At the heart of our method, and corresponding analysis, is a novel way to
compute decomposition-invariant away-steps. While our theoretical guarantees do
not apply to any polytope, they apply to several important structured polytopes
that capture central concepts such as paths in graphs, perfect matchings in
bipartite graphs, marginal distributions that arise in structured prediction
tasks, and more. Our theoretical findings are complemented by empirical
evidence which shows that our method delivers state-of-the-art performance.
|
TensorLog: A Differentiable Deductive Database | Large knowledge bases (KBs) are useful in many tasks, but it is unclear how
to integrate this sort of knowledge into "deep" gradient-based learning
systems. To address this problem, we describe a probabilistic deductive
database, called TensorLog, in which reasoning uses a differentiable process.
In TensorLog, each clause in a logical theory is first converted into certain
type of factor graph. Then, for each type of query to the factor graph, the
message-passing steps required to perform belief propagation (BP) are
"unrolled" into a function, which is differentiable. We show that these
functions can be composed recursively to perform inference in non-trivial
logical theories containing multiple interrelated clauses and predicates. Both
compilation and inference in TensorLog are efficient: compilation is linear in
theory size and proof depth, and inference is linear in database size and the
number of message-passing steps used in BP. We also present experimental
results with TensorLog and discuss its relationship to other first-order
probabilistic logics.
|
Functional Hashing for Compressing Neural Networks | As the complexity of deep neural networks (DNNs) trend to grow to absorb the
increasing sizes of data, memory and energy consumption has been receiving more
and more attentions for industrial applications, especially on mobile devices.
This paper presents a novel structure based on functional hashing to compress
DNNs, namely FunHashNN. For each entry in a deep net, FunHashNN uses multiple
low-cost hash functions to fetch values in the compression space, and then
employs a small reconstruction network to recover that entry. The
reconstruction network is plugged into the whole network and trained jointly.
FunHashNN includes the recently proposed HashedNets as a degenerated case, and
benefits from larger value capacity and less reconstruction loss. We further
discuss extensions with dual space hashing and multi-hops. On several benchmark
datasets, FunHashNN demonstrates high compression ratios with little loss on
prediction accuracy.
|
DynaNewton - Accelerating Newton's Method for Machine Learning | Newton's method is a fundamental technique in optimization with quadratic
convergence within a neighborhood around the optimum. However reaching this
neighborhood is often slow and dominates the computational costs. We exploit
two properties specific to empirical risk minimization problems to accelerate
Newton's method, namely, subsampling training data and increasing strong
convexity through regularization. We propose a novel continuation method, where
we define a family of objectives over increasing sample sizes and with
decreasing regularization strength. Solutions on this path are tracked such
that the minimizer of the previous objective is guaranteed to be within the
quadratic convergence region of the next objective to be optimized. Thereby
every Newton iteration is guaranteed to achieve super-linear contractions with
regard to the chosen objective, which becomes a moving target. We provide a
theoretical analysis that motivates our algorithm, called DynaNewton, and
characterizes its speed of convergence. Experiments on a wide range of data
sets and problems consistently confirm the predicted computational savings.
|
Online Influence Maximization under Independent Cascade Model with
Semi-Bandit Feedback | We study the online influence maximization problem in social networks under
the independent cascade model. Specifically, we aim to learn the set of "best
influencers" in a social network online while repeatedly interacting with it.
We address the challenges of (i) combinatorial action space, since the number
of feasible influencer sets grows exponentially with the maximum number of
influencers, and (ii) limited feedback, since only the influenced portion of
the network is observed. Under a stochastic semi-bandit feedback, we propose
and analyze IMLinUCB, a computationally efficient UCB-based algorithm. Our
bounds on the cumulative regret are polynomial in all quantities of interest,
achieve near-optimal dependence on the number of interactions and reflect the
topology of the network and the activation probabilities of its edges, thereby
giving insights on the problem complexity. To the best of our knowledge, these
are the first such results. Our experiments show that in several representative
graph topologies, the regret of IMLinUCB scales as suggested by our upper
bounds. IMLinUCB permits linear generalization and thus is both statistically
and computationally suitable for large-scale problems. Our experiments also
show that IMLinUCB with linear generalization can lead to low regret in
real-world online influence maximization.
|
Make Workers Work Harder: Decoupled Asynchronous Proximal Stochastic
Gradient Descent | Asynchronous parallel optimization algorithms for solving large-scale machine
learning problems have drawn significant attention from academia to industry
recently. This paper proposes a novel algorithm, decoupled asynchronous
proximal stochastic gradient descent (DAP-SGD), to minimize an objective
function that is the composite of the average of multiple empirical losses and
a regularization term. Unlike the traditional asynchronous proximal stochastic
gradient descent (TAP-SGD) in which the master carries much of the computation
load, the proposed algorithm off-loads the majority of computation tasks from
the master to workers, and leaves the master to conduct simple addition
operations. This strategy yields an easy-to-parallelize algorithm, whose
performance is justified by theoretical convergence analyses. To be specific,
DAP-SGD achieves an $O(\log T/T)$ rate when the step-size is diminishing and an
ergodic $O(1/\sqrt{T})$ rate when the step-size is constant, where $T$ is the
number of total iterations.
|
Deep Transfer Learning with Joint Adaptation Networks | Deep networks have been successfully applied to learn transferable features
for adapting models from a source domain to a different target domain. In this
paper, we present joint adaptation networks (JAN), which learn a transfer
network by aligning the joint distributions of multiple domain-specific layers
across domains based on a joint maximum mean discrepancy (JMMD) criterion.
Adversarial training strategy is adopted to maximize JMMD such that the
distributions of the source and target domains are made more distinguishable.
Learning can be performed by stochastic gradient descent with the gradients
computed by back-propagation in linear-time. Experiments testify that our model
yields state of the art results on standard datasets.
|
Programming with a Differentiable Forth Interpreter | Given that in practice training data is scarce for all but a small set of
problems, a core question is how to incorporate prior knowledge into a model.
In this paper, we consider the case of prior procedural knowledge for neural
networks, such as knowing how a program should traverse a sequence, but not
what local actions should be performed at each step. To this end, we present an
end-to-end differentiable interpreter for the programming language Forth which
enables programmers to write program sketches with slots that can be filled
with behaviour trained from program input-output data. We can optimise this
behaviour directly through gradient descent techniques on user-specified
objectives, and also integrate the program into any larger neural computation
graph. We show empirically that our interpreter is able to effectively leverage
different levels of prior program structure and learn complex behaviours such
as sequence sorting and addition. When connected to outputs of an LSTM and
trained jointly, our interpreter achieves state-of-the-art accuracy for
end-to-end reasoning about quantities expressed in natural language stories.
|
Latent Tree Models for Hierarchical Topic Detection | We present a novel method for hierarchical topic detection where topics are
obtained by clustering documents in multiple ways. Specifically, we model
document collections using a class of graphical models called hierarchical
latent tree models (HLTMs). The variables at the bottom level of an HLTM are
observed binary variables that represent the presence/absence of words in a
document. The variables at other levels are binary latent variables, with those
at the lowest latent level representing word co-occurrence patterns and those
at higher levels representing co-occurrence of patterns at the level below.
Each latent variable gives a soft partition of the documents, and document
clusters in the partitions are interpreted as topics. Latent variables at high
levels of the hierarchy capture long-range word co-occurrence patterns and
hence give thematically more general topics, while those at low levels of the
hierarchy capture short-range word co-occurrence patterns and give thematically
more specific topics. Unlike LDA-based topic models, HLTMs do not refer to a
document generation process and use word variables instead of token variables.
They use a tree structure to model the relationships between topics and words,
which is conducive to the discovery of meaningful topics and topic hierarchies.
|
Gambler's Ruin Bandit Problem | In this paper, we propose a new multi-armed bandit problem called the
Gambler's Ruin Bandit Problem (GRBP). In the GRBP, the learner proceeds in a
sequence of rounds, where each round is a Markov Decision Process (MDP) with
two actions (arms): a continuation action that moves the learner randomly over
the state space around the current state; and a terminal action that moves the
learner directly into one of the two terminal states (goal and dead-end state).
The current round ends when a terminal state is reached, and the learner incurs
a positive reward only when the goal state is reached. The objective of the
learner is to maximize its long-term reward (expected number of times the goal
state is reached), without having any prior knowledge on the state transition
probabilities. We first prove a result on the form of the optimal policy for
the GRBP. Then, we define the regret of the learner with respect to an
omnipotent oracle, which acts optimally in each round, and prove that it
increases logarithmically over rounds. We also identify a condition under which
the learner's regret is bounded. A potential application of the GRBP is optimal
medical treatment assignment, in which the continuation action corresponds to a
conservative treatment and the terminal action corresponds to a risky treatment
such as surgery.
|
Bending the Curve: Improving the ROC Curve Through Error Redistribution | Classification performance is often not uniform over the data. Some areas in
the input space are easier to classify than others. Features that hold
information about the "difficulty" of the data may be non-discriminative and
are therefore disregarded in the classification process. We propose a
meta-learning approach where performance may be improved by post-processing.
This improvement is done by establishing a dynamic threshold on the
base-classifier results. Since the base-classifier is treated as a "black box"
the method presented can be used on any state of the art classifier in order to
try an improve its performance. We focus our attention on how to better control
the true-positive/false-positive trade-off known as the ROC curve. We propose
an algorithm for the derivation of optimal thresholds by redistributing the
error depending on features that hold information about difficulty. We
demonstrate the resulting benefit on both synthetic and real-life data.
|
Cross Domain Adaptation by Learning Partially Shared Classifiers and
Weighting Source Data Points in the Shared Subspaces | Transfer learning is a problem defined over two domains. These two domains
share the same feature space and class label space, but have significantly
different distributions. One domain has sufficient labels, named as source
domain, and the other domain has few labels, named as target do- main. The
problem is to learn a effective classifier for the target domain. In this
paper, we propose a novel transfer learning method for this problem by learning
a partially shared classifier for the target domain, and weighting the source
domain data points. We learn some shared subspaces for both the data points of
the two domains, and a shared classifier in the shared subspaces. We hope that
in the shared subspaces, the distributions of two domain can match each other
well, and to match the distributions, we weight the source domain data points
with different weighting factors. Moreover, we adapt the shared classifier to
each domain by learning different adaptation functions. To learn the subspace
transformation matrices, the classifier parameters, and the adaptation
parameters, we build a objective function with weighted clas- sification
errors, parameter regularization, local reconstruction regularization, and
distribution matching. This objective function is minimized by an itera- tive
algorithm. Experiments show its effectiveness over benchmark data sets,
including travel destination review data set, face expression data set, spam
email data set, etc.
|
Learning to Communicate with Deep Multi-Agent Reinforcement Learning | We consider the problem of multiple agents sensing and acting in environments
with the goal of maximising their shared utility. In these environments, agents
must learn communication protocols in order to share information that is needed
to solve the tasks. By embracing deep neural networks, we are able to
demonstrate end-to-end learning of protocols in complex environments inspired
by communication riddles and multi-agent computer vision problems with partial
observability. We propose two approaches for learning in these domains:
Reinforced Inter-Agent Learning (RIAL) and Differentiable Inter-Agent Learning
(DIAL). The former uses deep Q-learning, while the latter exploits the fact
that, during learning, agents can backpropagate error derivatives through
(noisy) communication channels. Hence, this approach uses centralised learning
but decentralised execution. Our experiments introduce new environments for
studying the learning of communication protocols and present a set of
engineering innovations that are essential for success in these domains.
|
Efficient Document Indexing Using Pivot Tree | We present a novel method for efficiently searching top-k neighbors for
documents represented in high dimensional space of terms based on the cosine
similarity. Mostly, documents are stored as bag-of-words tf-idf representation.
One of the most used ways of computing similarity between a pair of documents
is cosine similarity between the vector representations, but cosine similarity
is not a metric distance measure as it doesn't follow triangle inequality,
therefore most metric searching methods can not be applied directly. We propose
an efficient method for indexing documents using a pivot tree that leads to
efficient retrieval. We also study the relation between precision and
efficiency for the proposed method and compare it with a state of the art in
the area of document searching based on inner product.
|
Learning From Hidden Traits: Joint Factor Analysis and Latent Clustering | Dimensionality reduction techniques play an essential role in data analytics,
signal processing and machine learning. Dimensionality reduction is usually
performed in a preprocessing stage that is separate from subsequent data
analysis, such as clustering or classification. Finding reduced-dimension
representations that are well-suited for the intended task is more appealing.
This paper proposes a joint factor analysis and latent clustering framework,
which aims at learning cluster-aware low-dimensional representations of matrix
and tensor data. The proposed approach leverages matrix and tensor
factorization models that produce essentially unique latent representations of
the data to unravel latent cluster structure -- which is otherwise obscured
because of the freedom to apply an oblique transformation in latent space. At
the same time, latent cluster structure is used as prior information to enhance
the performance of factorization. Specific contributions include several
custom-built problem formulations, corresponding algorithms, and discussion of
associated convergence properties. Besides extensive simulations, real-world
datasets such as Reuters document data and MNIST image data are also employed
to showcase the effectiveness of the proposed approaches.
|
Factored Temporal Sigmoid Belief Networks for Sequence Learning | Deep conditional generative models are developed to simultaneously learn the
temporal dependencies of multiple sequences. The model is designed by
introducing a three-way weight tensor to capture the multiplicative
interactions between side information and sequences. The proposed model builds
on the Temporal Sigmoid Belief Network (TSBN), a sequential stack of Sigmoid
Belief Networks (SBNs). The transition matrices are further factored to reduce
the number of parameters and improve generalization. When side information is
not available, a general framework for semi-supervised learning based on the
proposed model is constituted, allowing robust sequence classification.
Experimental results show that the proposed approach achieves state-of-the-art
predictive and classification performance on sequential data, and has the
capacity to synthesize sequences, with controlled style transitioning and
blending.
|
A Rapid Pattern-Recognition Method for Driving Types Using
Clustering-Based Support Vector Machines | A rapid pattern-recognition approach to characterize driver's
curve-negotiating behavior is proposed. To shorten the recognition time and
improve the recognition of driving styles, a k-means clustering-based support
vector machine ( kMC-SVM) method is developed and used for classifying drivers
into two types: aggressive and moderate. First, vehicle speed and throttle
opening are treated as the feature parameters to reflect the driving styles.
Second, to discriminate driver curve-negotiating behaviors and reduce the
number of support vectors, the k-means clustering method is used to extract and
gather the two types of driving data and shorten the recognition time. Then,
based on the clustering results, a support vector machine approach is utilized
to generate the hyperplane for judging and predicting to which types the human
driver are subject. Lastly, to verify the validity of the kMC-SVM method, a
cross-validation experiment is designed and conducted. The research results
show that the $ k $MC-SVM is an effective method to classify driving styles
with a short time, compared with SVM method.
|
Inductive Bias of Deep Convolutional Networks through Pooling Geometry | Our formal understanding of the inductive bias that drives the success of
convolutional networks on computer vision tasks is limited. In particular, it
is unclear what makes hypotheses spaces born from convolution and pooling
operations so suitable for natural images. In this paper we study the ability
of convolutional networks to model correlations among regions of their input.
We theoretically analyze convolutional arithmetic circuits, and empirically
validate our findings on other types of convolutional networks as well.
Correlations are formalized through the notion of separation rank, which for a
given partition of the input, measures how far a function is from being
separable. We show that a polynomially sized deep network supports
exponentially high separation ranks for certain input partitions, while being
limited to polynomial separation ranks for others. The network's pooling
geometry effectively determines which input partitions are favored, thus serves
as a means for controlling the inductive bias. Contiguous pooling windows as
commonly employed in practice favor interleaved partitions over coarse ones,
orienting the inductive bias towards the statistics of natural images. Other
pooling schemes lead to different preferences, and this allows tailoring the
network to data that departs from the usual domain of natural imagery. In
addition to analyzing deep networks, we show that shallow ones support only
linear separation ranks, and by this gain insight into the benefit of functions
brought forth by depth - they are able to efficiently model strong correlation
under favored partitions of the input.
|
Active Nearest-Neighbor Learning in Metric Spaces | We propose a pool-based non-parametric active learning algorithm for general
metric spaces, called MArgin Regularized Metric Active Nearest Neighbor
(MARMANN), which outputs a nearest-neighbor classifier. We give prediction
error guarantees that depend on the noisy-margin properties of the input
sample, and are competitive with those obtained by previously proposed passive
learners. We prove that the label complexity of MARMANN is significantly lower
than that of any passive learner with similar error guarantees. MARMANN is
based on a generalized sample compression scheme, and a new label-efficient
active model-selection procedure.
|
Interpretable Distribution Features with Maximum Testing Power | Two semimetrics on probability distributions are proposed, given as the sum
of differences of expectations of analytic functions evaluated at spatial or
frequency locations (i.e, features). The features are chosen so as to maximize
the distinguishability of the distributions, by optimizing a lower bound on
test power for a statistical test using these features. The result is a
parsimonious and interpretable indication of how and where two distributions
differ locally. An empirical estimate of the test power criterion converges
with increasing sample size, ensuring the quality of the returned features. In
real-world benchmarks on high-dimensional text and image data, linear-time
tests using the proposed semimetrics achieve comparable performance to the
state-of-the-art quadratic-time maximum mean discrepancy test, while returning
human-interpretable features that explain the test results.
|
Nonnegative Matrix Factorization Requires Irrationality | Nonnegative matrix factorization (NMF) is the problem of decomposing a given
nonnegative $n \times m$ matrix $M$ into a product of a nonnegative $n \times
d$ matrix $W$ and a nonnegative $d \times m$ matrix $H$. A longstanding open
question, posed by Cohen and Rothblum in 1993, is whether a rational matrix $M$
always has an NMF of minimal inner dimension $d$ whose factors $W$ and $H$ are
also rational. We answer this question negatively, by exhibiting a matrix for
which $W$ and $H$ require irrational entries.
|
Smart broadcasting: Do you want to be seen? | Many users in online social networks are constantly trying to gain attention
from their followers by broadcasting posts to them. These broadcasters are
likely to gain greater attention if their posts can remain visible for a longer
period of time among their followers' most recent feeds. Then when to post? In
this paper, we study the problem of smart broadcasting using the framework of
temporal point processes, where we model users feeds and posts as discrete
events occurring in continuous time. Based on such continuous-time model, then
choosing a broadcasting strategy for a user becomes a problem of designing the
conditional intensity of her posting events. We derive a novel formula which
links this conditional intensity with the visibility of the user in her
followers' feeds. Furthermore, by exploiting this formula, we develop an
efficient convex optimization framework for the when-to-post problem. Our
method can find broadcasting strategies that reach a desired visibility level
with provable guarantees. We experimented with data gathered from Twitter, and
show that our framework can consistently make broadcasters' post more visible
than alternatives.
|
DLAU: A Scalable Deep Learning Accelerator Unit on FPGA | As the emerging field of machine learning, deep learning shows excellent
ability in solving complex learning problems. However, the size of the networks
becomes increasingly large scale due to the demands of the practical
applications, which poses significant challenge to construct a high performance
implementations of deep learning neural networks. In order to improve the
performance as well to maintain the low power cost, in this paper we design
DLAU, which is a scalable accelerator architecture for large-scale deep
learning networks using FPGA as the hardware prototype. The DLAU accelerator
employs three pipelined processing units to improve the throughput and utilizes
tile techniques to explore locality for deep learning applications.
Experimental results on the state-of-the-art Xilinx FPGA board demonstrate that
the DLAU accelerator is able to achieve up to 36.1x speedup comparing to the
Intel Core2 processors, with the power consumption at 234mW.
|
Fast Stochastic Methods for Nonsmooth Nonconvex Optimization | We analyze stochastic algorithms for optimizing nonconvex, nonsmooth
finite-sum problems, where the nonconvex part is smooth and the nonsmooth part
is convex. Surprisingly, unlike the smooth case, our knowledge of this
fundamental problem is very limited. For example, it is not known whether the
proximal stochastic gradient method with constant minibatch converges to a
stationary point. To tackle this issue, we develop fast stochastic algorithms
that provably converge to a stationary point for constant minibatches.
Furthermore, using a variant of these algorithms, we show provably faster
convergence than batch proximal gradient descent. Finally, we prove global
linear convergence rate for an interesting subclass of nonsmooth nonconvex
functions, that subsumes several recent works. This paper builds upon our
recent series of papers on fast stochastic methods for smooth nonconvex
optimization [22, 23], with a novel analysis for nonconvex and nonsmooth
functions.
|
Generative Choreography using Deep Learning | Recent advances in deep learning have enabled the extraction of high-level
features from raw sensor data which has opened up new possibilities in many
different fields, including computer generated choreography. In this paper we
present a system chor-rnn for generating novel choreographic material in the
nuanced choreographic language and style of an individual choreographer. It
also shows promising results in producing a higher level compositional
cohesion, rather than just generating sequences of movement. At the core of
chor-rnn is a deep recurrent neural network trained on raw motion capture data
and that can generate new dance sequences for a solo dancer. Chor-rnn can be
used for collaborative human-machine choreography or as a creative catalyst,
serving as inspiration for a choreographer.
|
An Information Criterion for Inferring Coupling in Distributed Dynamical
Systems | The behaviour of many real-world phenomena can be modelled by nonlinear
dynamical systems whereby a latent system state is observed through a filter.
We are interested in interacting subsystems of this form, which we model by a
set of coupled maps as a synchronous update graph dynamical systems.
Specifically, we study the structure learning problem for spatially distributed
dynamical systems coupled via a directed acyclic graph. Unlike established
structure learning procedures that find locally maximum posterior probabilities
of a network structure containing latent variables, our work exploits the
properties of dynamical systems to compute globally optimal approximations of
these distributions. We arrive at this result by the use of time delay
embedding theorems. Taking an information-theoretic perspective, we show that
the log-likelihood has an intuitive interpretation in terms of information
transfer.
|
A Sub-Quadratic Exact Medoid Algorithm | We present a new algorithm, trimed, for obtaining the medoid of a set, that
is the element of the set which minimises the mean distance to all other
elements. The algorithm is shown to have, under certain assumptions, expected
run time O(N^(3/2)) in R^d where N is the set size, making it the first
sub-quadratic exact medoid algorithm for d>1. Experiments show that it performs
very well on spatial network data, frequently requiring two orders of magnitude
fewer distance calculations than state-of-the-art approximate algorithms. As an
application, we show how trimed can be used as a component in an accelerated
K-medoids algorithm, and then how it can be relaxed to obtain further
computational gains with only a minor loss in cluster quality.
|
Semi-Supervised Classification Based on Classification from Positive and
Unlabeled Data | Most of the semi-supervised classification methods developed so far use
unlabeled data for regularization purposes under particular distributional
assumptions such as the cluster assumption. In contrast, recently developed
methods of classification from positive and unlabeled data (PU classification)
use unlabeled data for risk evaluation, i.e., label information is directly
extracted from unlabeled data. In this paper, we extend PU classification to
also incorporate negative data and propose a novel semi-supervised
classification approach. We establish generalization error bounds for our novel
methods and show that the bounds decrease with respect to the number of
unlabeled data without the distributional assumptions that are required in
existing semi-supervised classification methods. Through experiments, we
demonstrate the usefulness of the proposed methods.
|
A Riemannian gossip approach to decentralized matrix completion | In this paper, we propose novel gossip algorithms for the low-rank
decentralized matrix completion problem. The proposed approach is on the
Riemannian Grassmann manifold that allows local matrix completion by different
agents while achieving asymptotic consensus on the global low-rank factors. The
resulting approach is scalable and parallelizable. Our numerical experiments
show the good performance of the proposed algorithms on various benchmarks.
|
DP-EM: Differentially Private Expectation Maximization | The iterative nature of the expectation maximization (EM) algorithm presents
a challenge for privacy-preserving estimation, as each iteration increases the
amount of noise needed. We propose a practical private EM algorithm that
overcomes this challenge using two innovations: (1) a novel moment perturbation
formulation for differentially private EM (DP-EM), and (2) the use of two
recently developed composition methods to bound the privacy "cost" of multiple
EM iterations: the moments accountant (MA) and zero-mean concentrated
differential privacy (zCDP). Both MA and zCDP bound the moment generating
function of the privacy loss random variable and achieve a refined tail bound,
which effectively decrease the amount of additive noise. We present empirical
results showing the benefits of our approach, as well as similar performance
between these two composition methods in the DP-EM setting for Gaussian mixture
models. Our approach can be readily extended to many iterative learning
algorithms, opening up various exciting future directions.
|
Online Learning with Feedback Graphs Without the Graphs | We study an online learning framework introduced by Mannor and Shamir (2011)
in which the feedback is specified by a graph, in a setting where the graph may
vary from round to round and is \emph{never fully revealed} to the learner. We
show a large gap between the adversarial and the stochastic cases. In the
adversarial case, we prove that even for dense feedback graphs, the learner
cannot improve upon a trivial regret bound obtained by ignoring any additional
feedback besides her own loss. In contrast, in the stochastic case we give an
algorithm that achieves $\widetilde \Theta(\sqrt{\alpha T})$ regret over $T$
rounds, provided that the independence numbers of the hidden feedback graphs
are at most $\alpha$. We also extend our results to a more general feedback
model, in which the learner does not necessarily observe her own loss, and show
that, even in simple cases, concealing the feedback graphs might render a
learnable problem unlearnable.
|
Collaborative Filtering with Side Information: a Gaussian Process
Perspective | We tackle the problem of collaborative filtering (CF) with side information,
through the lens of Gaussian Process (GP) regression. Driven by the idea of
using the kernel to explicitly model user-item similarities, we formulate the
GP in a way that allows the incorporation of low-rank matrix factorisation,
arriving at our model, the Tucker Gaussian Process (TGP). Consequently, TGP
generalises classical Bayesian matrix factorisation models, and goes beyond
them to give a natural and elegant method for incorporating side information,
giving enhanced predictive performance for CF problems. Moreover we show that
it is a novel model for regression, especially well-suited to grid-structured
data and problems where the dependence on covariates is close to being
separable.
|
Convergence Analysis for Rectangular Matrix Completion Using
Burer-Monteiro Factorization and Gradient Descent | We address the rectangular matrix completion problem by lifting the unknown
matrix to a positive semidefinite matrix in higher dimension, and optimizing a
nonconvex objective over the semidefinite factor using a simple gradient
descent scheme. With $O( \mu r^2 \kappa^2 n \max(\mu, \log n))$ random
observations of a $n_1 \times n_2$ $\mu$-incoherent matrix of rank $r$ and
condition number $\kappa$, where $n = \max(n_1, n_2)$, the algorithm linearly
converges to the global optimum with high probability.
|
Bayesian Model Selection of Stochastic Block Models | A central problem in analyzing networks is partitioning them into modules or
communities. One of the best tools for this is the stochastic block model,
which clusters vertices into blocks with statistically homogeneous pattern of
links. Despite its flexibility and popularity, there has been a lack of
principled statistical model selection criteria for the stochastic block model.
Here we propose a Bayesian framework for choosing the number of blocks as well
as comparing it to the more elaborate degree- corrected block models,
ultimately leading to a universal model selection framework capable of
comparing multiple modeling combinations. We will also investigate its
connection to the minimum description length principle.
|
On Restricted Nonnegative Matrix Factorization | Nonnegative matrix factorization (NMF) is the problem of decomposing a given
nonnegative $n \times m$ matrix $M$ into a product of a nonnegative $n \times
d$ matrix $W$ and a nonnegative $d \times m$ matrix $H$. Restricted NMF
requires in addition that the column spaces of $M$ and $W$ coincide. Finding
the minimal inner dimension $d$ is known to be NP-hard, both for NMF and
restricted NMF. We show that restricted NMF is closely related to a question
about the nature of minimal probabilistic automata, posed by Paz in his seminal
1971 textbook. We use this connection to answer Paz's question negatively, thus
falsifying a positive answer claimed in 1974. Furthermore, we investigate
whether a rational matrix $M$ always has a restricted NMF of minimal inner
dimension whose factors $W$ and $H$ are also rational. We show that this holds
for matrices $M$ of rank at most $3$ and we exhibit a rank-$4$ matrix for which
$W$ and $H$ require irrational entries.
|
A Unifying Framework for Gaussian Process Pseudo-Point Approximations
using Power Expectation Propagation | Gaussian processes (GPs) are flexible distributions over functions that
enable high-level assumptions about unknown functions to be encoded in a
parsimonious, flexible and general way. Although elegant, the application of
GPs is limited by computational and analytical intractabilities that arise when
data are sufficiently numerous or when employing non-Gaussian models.
Consequently, a wealth of GP approximation schemes have been developed over the
last 15 years to address these key limitations. Many of these schemes employ a
small set of pseudo data points to summarise the actual data. In this paper, we
develop a new pseudo-point approximation framework using Power Expectation
Propagation (Power EP) that unifies a large number of these pseudo-point
approximations. Unlike much of the previous venerable work in this area, the
new framework is built on standard methods for approximate inference
(variational free-energy, EP and Power EP methods) rather than employing
approximations to the probabilistic generative model itself. In this way, all
of approximation is performed at `inference time' rather than at `modelling
time' resolving awkward philosophical and empirical questions that trouble
previous approaches. Crucially, we demonstrate that the new framework includes
new pseudo-point approximation methods that outperform current approaches on
regression and classification tasks.
|
Learning Sensor Multiplexing Design through Back-propagation | Recent progress on many imaging and vision tasks has been driven by the use
of deep feed-forward neural networks, which are trained by propagating
gradients of a loss defined on the final output, back through the network up to
the first layer that operates directly on the image. We propose
back-propagating one step further---to learn camera sensor designs jointly with
networks that carry out inference on the images they capture. In this paper, we
specifically consider the design and inference problems in a typical color
camera---where the sensor is able to measure only one color channel at each
pixel location, and computational inference is required to reconstruct a full
color image. We learn the camera sensor's color multiplexing pattern by
encoding it as layer whose learnable weights determine which color channel,
from among a fixed set, will be measured at each location. These weights are
jointly trained with those of a reconstruction network that operates on the
corresponding sensor measurements to produce a full color image. Our network
achieves significant improvements in accuracy over the traditional Bayer
pattern used in most color cameras. It automatically learns to employ a sparse
color measurement approach similar to that of a recent design, and moreover,
improves upon that design by learning an optimal layout for these measurements.
|
Fast Bayesian Optimization of Machine Learning Hyperparameters on Large
Datasets | Bayesian optimization has become a successful tool for hyperparameter
optimization of machine learning algorithms, such as support vector machines or
deep neural networks. Despite its success, for large datasets, training and
validating a single configuration often takes hours, days, or even weeks, which
limits the achievable performance. To accelerate hyperparameter optimization,
we propose a generative model for the validation error as a function of
training set size, which is learned during the optimization process and allows
exploration of preliminary configurations on small subsets, by extrapolating to
the full dataset. We construct a Bayesian optimization procedure, dubbed
Fabolas, which models loss and training time as a function of dataset size and
automatically trades off high information gain about the global optimum against
computational cost. Experiments optimizing support vector machines and deep
neural networks show that Fabolas often finds high-quality solutions 10 to 100
times faster than other state-of-the-art Bayesian optimization methods or the
recently proposed bandit strategy Hyperband.
|
A note on the expected minimum error probability in equientropic
channels | While the channel capacity reflects a theoretical upper bound on the
achievable information transmission rate in the limit of infinitely many bits,
it does not characterise the information transfer of a given encoding routine
with finitely many bits. In this note, we characterise the quality of a code
(i. e. a given encoding routine) by an upper bound on the expected minimum
error probability that can be achieved when using this code. We show that for
equientropic channels this upper bound is minimal for codes with maximal
marginal entropy. As an instructive example we show for the additive white
Gaussian noise (AWGN) channel that random coding---also a capacity achieving
code---indeed maximises the marginal entropy in the limit of infinite messages.
|
Deep Learning without Poor Local Minima | In this paper, we prove a conjecture published in 1989 and also partially
address an open problem announced at the Conference on Learning Theory (COLT)
2015. With no unrealistic assumption, we first prove the following statements
for the squared loss function of deep linear neural networks with any depth and
any widths: 1) the function is non-convex and non-concave, 2) every local
minimum is a global minimum, 3) every critical point that is not a global
minimum is a saddle point, and 4) there exist "bad" saddle points (where the
Hessian has no negative eigenvalue) for the deeper networks (with more than
three layers), whereas there is no bad saddle point for the shallow networks
(with three layers). Moreover, for deep nonlinear neural networks, we prove the
same four statements via a reduction to a deep linear model under the
independence assumption adopted from recent work. As a result, we present an
instance, for which we can answer the following question: how difficult is it
to directly train a deep model in theory? It is more difficult than the
classical machine learning models (because of the non-convexity), but not too
difficult (because of the nonexistence of poor local minima). Furthermore, the
mathematically proven existence of bad saddle points for deeper models would
suggest a possible open problem. We note that even though we have advanced the
theoretical foundations of deep learning and non-convex optimization, there is
still a gap between theory and practice.
|
Learning and Policy Search in Stochastic Dynamical Systems with Bayesian
Neural Networks | We present an algorithm for model-based reinforcement learning that combines
Bayesian neural networks (BNNs) with random roll-outs and stochastic
optimization for policy learning. The BNNs are trained by minimizing
$\alpha$-divergences, allowing us to capture complicated statistical patterns
in the transition dynamics, e.g. multi-modality and heteroskedasticity, which
are usually missed by other common modeling approaches. We illustrate the
performance of our method by solving a challenging benchmark where model-based
approaches usually fail and by obtaining promising results in a real-world
scenario for controlling a gas turbine.
|
Towards Multi-Agent Communication-Based Language Learning | We propose an interactive multimodal framework for language learning. Instead
of being passively exposed to large amounts of natural text, our learners
(implemented as feed-forward neural networks) engage in cooperative referential
games starting from a tabula rasa setup, and thus develop their own language
from the need to communicate in order to succeed at the game. Preliminary
experiments provide promising results, but also suggest that it is important to
ensure that agents trained in this way do not develop an adhoc communication
code only effective for the game they are playing
|
Fairness in Learning: Classic and Contextual Bandits | We introduce the study of fairness in multi-armed bandit problems. Our
fairness definition can be interpreted as demanding that given a pool of
applicants (say, for college admission or mortgages), a worse applicant is
never favored over a better one, despite a learning algorithm's uncertainty
over the true payoffs. We prove results of two types.
First, in the important special case of the classic stochastic bandits
problem (i.e., in which there are no contexts), we provide a provably fair
algorithm based on "chained" confidence intervals, and provide a cumulative
regret bound with a cubic dependence on the number of arms. We further show
that any fair algorithm must have such a dependence. When combined with regret
bounds for standard non-fair algorithms such as UCB, this proves a strong
separation between fair and unfair learning, which extends to the general
contextual case.
In the general contextual case, we prove a tight connection between fairness
and the KWIK (Knows What It Knows) learning model: a KWIK algorithm for a class
of functions can be transformed into a provably fair contextual bandit
algorithm, and conversely any fair contextual bandit algorithm can be
transformed into a KWIK learning algorithm. This tight connection allows us to
provide a provably fair algorithm for the linear contextual bandit problem with
a polynomial dependence on the dimension, and to show (for a different class of
functions) a worst-case exponential gap in regret between fair and non-fair
learning algorithms
|
Actively Learning Hemimetrics with Applications to Eliciting User
Preferences | Motivated by an application of eliciting users' preferences, we investigate
the problem of learning hemimetrics, i.e., pairwise distances among a set of
$n$ items that satisfy triangle inequalities and non-negativity constraints. In
our application, the (asymmetric) distances quantify private costs a user
incurs when substituting one item by another. We aim to learn these distances
(costs) by asking the users whether they are willing to switch from one item to
another for a given incentive offer. Without exploiting structural constraints
of the hemimetric polytope, learning the distances between each pair of items
requires $\Theta(n^2)$ queries. We propose an active learning algorithm that
substantially reduces this sample complexity by exploiting the structural
constraints on the version space of hemimetrics. Our proposed algorithm
achieves provably-optimal sample complexity for various instances of the task.
For example, when the items are embedded into $K$ tight clusters, the sample
complexity of our algorithm reduces to $O(n K)$. Extensive experiments on a
restaurant recommendation data set support the conclusions of our theoretical
analysis.
|
On Optimality Conditions for Auto-Encoder Signal Recovery | Auto-Encoders are unsupervised models that aim to learn patterns from
observed data by minimizing a reconstruction cost. The useful representations
learned are often found to be sparse and distributed. On the other hand,
compressed sensing and sparse coding assume a data generating process, where
the observed data is generated from some true latent signal source, and try to
recover the corresponding signal from measurements. Looking at auto-encoders
from this \textit{signal recovery perspective} enables us to have a more
coherent view of these techniques. In this paper, in particular, we show that
the \textit{true} hidden representation can be approximately recovered if the
weight matrices are highly incoherent with unit $ \ell^{2} $ row length and the
bias vectors takes the value (approximately) equal to the negative of the data
mean. The recovery also becomes more and more accurate as the sparsity in
hidden signals increases. Additionally, we empirically demonstrate that
auto-encoders are capable of recovering the data generating dictionary when
only data samples are given.
|
Wide Residual Networks | Deep residual networks were shown to be able to scale up to thousands of
layers and still have improving performance. However, each fraction of a
percent of improved accuracy costs nearly doubling the number of layers, and so
training very deep residual networks has a problem of diminishing feature
reuse, which makes these networks very slow to train. To tackle these problems,
in this paper we conduct a detailed experimental study on the architecture of
ResNet blocks, based on which we propose a novel architecture where we decrease
depth and increase width of residual networks. We call the resulting network
structures wide residual networks (WRNs) and show that these are far superior
over their commonly used thin and very deep counterparts. For example, we
demonstrate that even a simple 16-layer-deep wide residual network outperforms
in accuracy and efficiency all previous deep residual networks, including
thousand-layer-deep networks, achieving new state-of-the-art results on CIFAR,
SVHN, COCO, and significant improvements on ImageNet. Our code and models are
available at https://github.com/szagoruyko/wide-residual-networks
|
Riemannian SVRG: Fast Stochastic Optimization on Riemannian Manifolds | We study optimization of finite sums of geodesically smooth functions on
Riemannian manifolds. Although variance reduction techniques for optimizing
finite-sums have witnessed tremendous attention in the recent years, existing
work is limited to vector space problems. We introduce Riemannian SVRG (RSVRG),
a new variance reduced Riemannian optimization method. We analyze RSVRG for
both geodesically convex and nonconvex (smooth) functions. Our analysis reveals
that RSVRG inherits advantages of the usual SVRG method, but with factors
depending on curvature of the manifold that influence its convergence. To our
knowledge, RSVRG is the first provably fast stochastic Riemannian method.
Moreover, our paper presents the first non-asymptotic complexity analysis
(novel even for the batch setting) for nonconvex Riemannian optimization. Our
results have several implications; for instance, they offer a Riemannian
perspective on variance reduced PCA, which promises a short, transparent
convergence analysis.
|
Backprop KF: Learning Discriminative Deterministic State Estimators | Generative state estimators based on probabilistic filters and smoothers are
one of the most popular classes of state estimators for robots and autonomous
vehicles. However, generative models have limited capacity to handle rich
sensory observations, such as camera images, since they must model the entire
distribution over sensor readings. Discriminative models do not suffer from
this limitation, but are typically more complex to train as latent variable
models for state estimation. We present an alternative approach where the
parameters of the latent state distribution are directly optimized as a
deterministic computation graph, resulting in a simple and effective gradient
descent algorithm for training discriminative state estimators. We show that
this procedure can be used to train state estimators that use complex input,
such as raw camera images, which must be processed using expressive nonlinear
function approximators such as convolutional neural networks. Our model can be
viewed as a type of recurrent neural network, and the connection to
probabilistic filtering allows us to design a network architecture that is
particularly well suited for state estimation. We evaluate our approach on
synthetic tracking task with raw image inputs and on the visual odometry task
in the KITTI dataset. The results show significant improvement over both
standard generative approaches and regular recurrent neural networks.
|
Path-Normalized Optimization of Recurrent Neural Networks with ReLU
Activations | We investigate the parameter-space geometry of recurrent neural networks
(RNNs), and develop an adaptation of path-SGD optimization method, attuned to
this geometry, that can learn plain RNNs with ReLU activations. On several
datasets that require capturing long-term dependency structure, we show that
path-SGD can significantly improve trainability of ReLU RNNs compared to RNNs
trained with SGD, even with various recently suggested initialization schemes.
|
Genetic Architect: Discovering Genomic Structure with Learned Neural
Architectures | Each human genome is a 3 billion base pair set of encoding instructions.
Decoding the genome using deep learning fundamentally differs from most tasks,
as we do not know the full structure of the data and therefore cannot design
architectures to suit it. As such, architectures that fit the structure of
genomics should be learned not prescribed. Here, we develop a novel search
algorithm, applicable across domains, that discovers an optimal architecture
which simultaneously learns general genomic patterns and identifies the most
important sequence motifs in predicting functional genomic outcomes. The
architectures we find using this algorithm succeed at using only RNA expression
data to predict gene regulatory structure, learn human-interpretable
visualizations of key sequence motifs, and surpass state-of-the-art results on
benchmark genomics challenges.
|
Unsupervised Learning for Physical Interaction through Video Prediction | A core challenge for an agent learning to interact with the world is to
predict how its actions affect objects in its environment. Many existing
methods for learning the dynamics of physical interactions require labeled
object information. However, to scale real-world interaction learning to a
variety of scenes and objects, acquiring labeled data becomes increasingly
impractical. To learn about physical object motion without labels, we develop
an action-conditioned video prediction model that explicitly models pixel
motion, by predicting a distribution over pixel motion from previous frames.
Because our model explicitly predicts motion, it is partially invariant to
object appearance, enabling it to generalize to previously unseen objects. To
explore video prediction for real-world interactive agents, we also introduce a
dataset of 59,000 robot interactions involving pushing motions, including a
test set with novel objects. In this dataset, accurate prediction of videos
conditioned on the robot's future actions amounts to learning a "visual
imagination" of different futures based on different courses of action. Our
experiments show that our proposed method produces more accurate video
predictions both quantitatively and qualitatively, when compared to prior
methods.
|
Pure Exploration of Multi-armed Bandit Under Matroid Constraints | We study the pure exploration problem subject to a matroid constraint
(Best-Basis) in a stochastic multi-armed bandit game. In a Best-Basis instance,
we are given $n$ stochastic arms with unknown reward distributions, as well as
a matroid $\mathcal{M}$ over the arms. Let the weight of an arm be the mean of
its reward distribution. Our goal is to identify a basis of $\mathcal{M}$ with
the maximum total weight, using as few samples as possible.
The problem is a significant generalization of the best arm identification
problem and the top-$k$ arm identification problem, which have attracted
significant attentions in recent years. We study both the exact and PAC
versions of Best-Basis, and provide algorithms with nearly-optimal sample
complexities for these versions. Our results generalize and/or improve on
several previous results for the top-$k$ arm identification problem and the
combinatorial pure exploration problem when the combinatorial constraint is a
matroid.
|
Kernel-based Reconstruction of Graph Signals | A number of applications in engineering, social sciences, physics, and
biology involve inference over networks. In this context, graph signals are
widely encountered as descriptors of vertex attributes or features in
graph-structured data. Estimating such signals in all vertices given noisy
observations of their values on a subset of vertices has been extensively
analyzed in the literature of signal processing on graphs (SPoG). This paper
advocates kernel regression as a framework generalizing popular SPoG modeling
and reconstruction and expanding their capabilities. Formulating signal
reconstruction as a regression task on reproducing kernel Hilbert spaces of
graph signals permeates benefits from statistical learning, offers fresh
insights, and allows for estimators to leverage richer forms of prior
information than existing alternatives. A number of SPoG notions such as
bandlimitedness, graph filters, and the graph Fourier transform are naturally
accommodated in the kernel framework. Additionally, this paper capitalizes on
the so-called representer theorem to devise simpler versions of existing
Thikhonov regularized estimators, and offers a novel probabilistic
interpretation of kernel methods on graphs based on graphical models. Motivated
by the challenges of selecting the bandwidth parameter in SPoG estimators or
the kernel map in kernel-based methods, the present paper further proposes two
multi-kernel approaches with complementary strengths. Whereas the first enables
estimation of the unknown bandwidth of bandlimited signals, the second allows
for efficient graph filter selection. Numerical tests with synthetic as well as
real data demonstrate the merits of the proposed methods relative to
state-of-the-art alternatives.
|
Global Optimality of Local Search for Low Rank Matrix Recovery | We show that there are no spurious local minima in the non-convex factorized
parametrization of low-rank matrix recovery from incoherent linear
measurements. With noisy measurements we show all local minima are very close
to a global optimum. Together with a curvature bound at saddle points, this
yields a polynomial time global convergence guarantee for stochastic gradient
descent {\em from random initialization}.
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.