title
stringlengths 7
246
| abstract
stringlengths 6
3.31k
|
|---|---|
Asymmetrically Weighted CCA And Hierarchical Kernel Sentence Embedding
For Image & Text Retrieval | Joint modeling of language and vision has been drawing increasing interest. A
multimodal data representation allowing for bidirectional retrieval of images
by sentences and vice versa is a key aspect. In this paper we present three
contributions in canonical correlation analysis (CCA) based multimodal
retrieval. Firstly, we show that an asymmetric weighting of the canonical
weights, while achieving a cross view mapping from the search to the query
space, improves the retrieval performance. Secondly, we devise a
computationally efficient model selection, crucial to generalization and
stability, in the framework of the Bj\"ork Golub algorithm for regularized CCA
via spectral filtering. Finally, we introduce a Hierarchical Kernel Sentence
Embedding (HKSE) that approximates Kernel CCA for a special similarity kernel
between distribution of words embedded in a vector space. State of the art
results are obtained on MSCOCO and Flickr benchmarks when these three
techniques are used in conjunction.
|
Faster method for Deep Belief Network based Object classification using
DWT | A Deep Belief Network (DBN) requires large, multiple hidden layers with high
number of hidden units to learn good features from the raw pixels of large
images. This implies more training time as well as computational complexity. By
integrating DBN with Discrete Wavelet Transform (DWT), both training time and
computational complexity can be reduced. The low resolution images obtained
after application of DWT are used to train multiple DBNs. The results obtained
from these DBNs are combined using a weighted voting algorithm. The performance
of this method is found to be competent and faster in comparison with that of
traditional DBNs.
|
Neural Programmer-Interpreters | We propose the neural programmer-interpreter (NPI): a recurrent and
compositional neural network that learns to represent and execute programs. NPI
has three learnable components: a task-agnostic recurrent core, a persistent
key-value program memory, and domain-specific encoders that enable a single NPI
to operate in multiple perceptually diverse environments with distinct
affordances. By learning to compose lower-level programs to express
higher-level programs, NPI reduces sample complexity and increases
generalization ability compared to sequence-to-sequence LSTMs. The program
memory allows efficient learning of additional tasks by building on existing
programs. NPI can also harness the environment (e.g. a scratch pad with
read-write pointers) to cache intermediate results of computation, lessening
the long-term memory burden on recurrent hidden units. In this work we train
the NPI with fully-supervised execution traces; each program has example
sequences of calls to the immediate subprograms conditioned on the input.
Rather than training on a huge number of relatively weak labels, NPI learns
from a small number of rich examples. We demonstrate the capability of our
model to learn several types of compositional programs: addition, sorting, and
canonicalizing 3D models. Furthermore, a single NPI learns to execute these
programs and all 21 associated subprograms.
|
Density Modeling of Images using a Generalized Normalization
Transformation | We introduce a parametric nonlinear transformation that is well-suited for
Gaussianizing data from natural images. The data are linearly transformed, and
each component is then normalized by a pooled activity measure, computed by
exponentiating a weighted sum of rectified and exponentiated components and a
constant. We optimize the parameters of the full transformation (linear
transform, exponents, weights, constant) over a database of natural images,
directly minimizing the negentropy of the responses. The optimized
transformation substantially Gaussianizes the data, achieving a significantly
smaller mutual information between transformed components than alternative
methods including ICA and radial Gaussianization. The transformation is
differentiable and can be efficiently inverted, and thus induces a density
model on images. We show that samples of this model are visually similar to
samples of natural image patches. We demonstrate the use of the model as a
prior probability density that can be used to remove additive noise. Finally,
we show that the transformation can be cascaded, with each layer optimized
using the same Gaussianization objective, thus offering an unsupervised method
of optimizing a deep network architecture.
|
Foveation-based Mechanisms Alleviate Adversarial Examples | We show that adversarial examples, i.e., the visually imperceptible
perturbations that result in Convolutional Neural Networks (CNNs) fail, can be
alleviated with a mechanism based on foveations---applying the CNN in different
image regions. To see this, first, we report results in ImageNet that lead to a
revision of the hypothesis that adversarial perturbations are a consequence of
CNNs acting as a linear classifier: CNNs act locally linearly to changes in the
image regions with objects recognized by the CNN, and in other regions the CNN
may act non-linearly. Then, we corroborate that when the neural responses are
linear, applying the foveation mechanism to the adversarial example tends to
significantly reduce the effect of the perturbation. This is because,
hypothetically, the CNNs for ImageNet are robust to changes of scale and
translation of the object produced by the foveation, but this property does not
generalize to transformations of the perturbation. As a result, the accuracy
after a foveation is almost the same as the accuracy of the CNN without the
adversarial perturbation, even if the adversarial perturbation is calculated
taking into account a foveation.
|
Policy Distillation | Policies for complex visual tasks have been successfully learned with deep
reinforcement learning, using an approach called deep Q-networks (DQN), but
relatively large (task-specific) networks and extensive training are needed to
achieve good performance. In this work, we present a novel method called policy
distillation that can be used to extract the policy of a reinforcement learning
agent and train a new network that performs at the expert level while being
dramatically smaller and more efficient. Furthermore, the same method can be
used to consolidate multiple task-specific policies into a single policy. We
demonstrate these claims using the Atari domain and show that the multi-task
distilled agent outperforms the single-task teachers as well as a
jointly-trained DQN agent.
|
Conditional Computation in Neural Networks for faster models | Deep learning has become the state-of-art tool in many applications, but the
evaluation and training of deep models can be time-consuming and
computationally expensive. The conditional computation approach has been
proposed to tackle this problem (Bengio et al., 2013; Davis & Arel, 2013). It
operates by selectively activating only parts of the network at a time. In this
paper, we use reinforcement learning as a tool to optimize conditional
computation policies. More specifically, we cast the problem of learning
activation-dependent policies for dropping out blocks of units as a
reinforcement learning problem. We propose a learning scheme motivated by
computation speed, capturing the idea of wanting to have parsimonious
activations while maintaining prediction accuracy. We apply a policy gradient
algorithm for learning policies that optimize this loss function and propose a
regularization mechanism that encourages diversification of the dropout policy.
We present encouraging empirical results showing that this approach improves
the speed of computation without impacting the quality of the approximation.
|
Alternative structures for character-level RNNs | Recurrent neural networks are convenient and efficient models for language
modeling. However, when applied on the level of characters instead of words,
they suffer from several problems. In order to successfully model long-term
dependencies, the hidden representation needs to be large. This in turn implies
higher computational costs, which can become prohibitive in practice. We
propose two alternative structural modifications to the classical RNN model.
The first one consists on conditioning the character level representation on
the previous word representation. The other one uses the character history to
condition the output probability. We evaluate the performance of the two
proposed modifications on challenging, multi-lingual real world data.
|
Robust Convolutional Neural Networks under Adversarial Noise | Recent studies have shown that Convolutional Neural Networks (CNNs) are
vulnerable to a small perturbation of input called "adversarial examples". In
this work, we propose a new feedforward CNN that improves robustness in the
presence of adversarial noise. Our model uses stochastic additive noise added
to the input image and to the CNN models. The proposed model operates in
conjunction with a CNN trained with either standard or adversarial objective
function. In particular, convolution, max-pooling, and ReLU layers are modified
to benefit from the noise model. Our feedforward model is parameterized by only
a mean and variance per pixel which simplifies computations and makes our
method scalable to a deep architecture. From CIFAR-10 and ImageNet test, the
proposed model outperforms other methods and the improvement is more evident
for difficult classification tasks or stronger adversarial noise.
|
Spatio-temporal video autoencoder with differentiable memory | We describe a new spatio-temporal video autoencoder, based on a classic
spatial image autoencoder and a novel nested temporal autoencoder. The temporal
encoder is represented by a differentiable visual memory composed of
convolutional long short-term memory (LSTM) cells that integrate changes over
time. Here we target motion changes and use as temporal decoder a robust
optical flow prediction module together with an image sampler serving as
built-in feedback loop. The architecture is end-to-end differentiable. At each
time step, the system receives as input a video frame, predicts the optical
flow based on the current observation and the LSTM memory state as a dense
transformation map, and applies it to the current frame to generate the next
frame. By minimising the reconstruction error between the predicted next frame
and the corresponding ground truth next frame, we train the whole system to
extract features useful for motion estimation without any supervision effort.
We present one direct application of the proposed framework in
weakly-supervised semantic segmentation of videos through label propagation
using optical flow.
|
Why M Heads are Better than One: Training a Diverse Ensemble of Deep
Networks | Convolutional Neural Networks have achieved state-of-the-art performance on a
wide range of tasks. Most benchmarks are led by ensembles of these powerful
learners, but ensembling is typically treated as a post-hoc procedure
implemented by averaging independently trained models with model variation
induced by bagging or random initialization. In this paper, we rigorously treat
ensembling as a first-class problem to explicitly address the question: what
are the best strategies to create an ensemble? We first compare a large number
of ensembling strategies, and then propose and evaluate novel strategies, such
as parameter sharing (through a new family of models we call TreeNets) as well
as training under ensemble-aware and diversity-encouraging losses. We
demonstrate that TreeNets can improve ensemble performance and that diverse
ensembles can be trained end-to-end under a unified loss, achieving
significantly higher "oracle" accuracies than classical ensembles.
|
Neural network-based clustering using pairwise constraints | This paper presents a neural network-based end-to-end clustering framework.
We design a novel strategy to utilize the contrastive criteria for pushing
data-forming clusters directly from raw data, in addition to learning a feature
embedding suitable for such clustering. The network is trained with weak
labels, specifically partial pairwise relationships between data instances. The
cluster assignments and their probabilities are then obtained at the output
layer by feed-forwarding the data. The framework has the interesting
characteristic that no cluster centers need to be explicitly specified, thus
the resulting cluster distribution is purely data-driven and no distance
metrics need to be predefined. The experiments show that the proposed approach
beats the conventional two-stage method (feature embedding with k-means) by a
significant margin. It also compares favorably to the performance of the
standard cross entropy loss for classification. Robustness analysis also shows
that the method is largely insensitive to the number of clusters. Specifically,
we show that the number of dominant clusters is close to the true number of
clusters even when a large k is used for clustering.
|
Manifold Regularized Discriminative Neural Networks | Unregularized deep neural networks (DNNs) can be easily overfit with a
limited sample size. We argue that this is mostly due to the disriminative
nature of DNNs which directly model the conditional probability (or score) of
labels given the input. The ignorance of input distribution makes DNNs
difficult to generalize to unseen data. Recent advances in regularization
techniques, such as pretraining and dropout, indicate that modeling input data
distribution (either explicitly or implicitly) greatly improves the
generalization ability of a DNN. In this work, we explore the manifold
hypothesis which assumes that instances within the same class lie in a smooth
manifold. We accordingly propose two simple regularizers to a standard
discriminative DNN. The first one, named Label-Aware Manifold Regularization,
assumes the availability of labels and penalizes large norms of the loss
function w.r.t. data points. The second one, named Label-Independent Manifold
Regularization, does not use label information and instead penalizes the
Frobenius norm of the Jacobian matrix of prediction scores w.r.t. data points,
which makes semi-supervised learning possible. We perform extensive control
experiments on fully supervised and semi-supervised tasks using the MNIST,
CIFAR10 and SVHN datasets and achieve excellent results.
|
Efficient Sum of Outer Products Dictionary Learning (SOUP-DIL) and Its
Application to Inverse Problems | The sparsity of signals in a transform domain or dictionary has been
exploited in applications such as compression, denoising and inverse problems.
More recently, data-driven adaptation of synthesis dictionaries has shown
promise compared to analytical dictionary models. However, dictionary learning
problems are typically non-convex and NP-hard, and the usual alternating
minimization approaches for these problems are often computationally expensive,
with the computations dominated by the NP-hard synthesis sparse coding step.
This paper exploits the ideas that drive algorithms such as K-SVD, and
investigates in detail efficient methods for aggregate sparsity penalized
dictionary learning by first approximating the data with a sum of sparse
rank-one matrices (outer products) and then using a block coordinate descent
approach to estimate the unknowns. The resulting block coordinate descent
algorithms involve efficient closed-form solutions. Furthermore, we consider
the problem of dictionary-blind image reconstruction, and propose novel and
efficient algorithms for adaptive image reconstruction using block coordinate
descent and sum of outer products methodologies. We provide a convergence study
of the algorithms for dictionary learning and dictionary-blind image
reconstruction. Our numerical experiments show the promising performance and
speed-ups provided by the proposed methods over previous schemes in sparse data
representation and compressed sensing-based image reconstruction.
|
Unsupervised Deep Embedding for Clustering Analysis | Clustering is central to many data-driven application domains and has been
studied extensively in terms of distance functions and grouping algorithms.
Relatively little work has focused on learning representations for clustering.
In this paper, we propose Deep Embedded Clustering (DEC), a method that
simultaneously learns feature representations and cluster assignments using
deep neural networks. DEC learns a mapping from the data space to a
lower-dimensional feature space in which it iteratively optimizes a clustering
objective. Our experimental evaluations on image and text corpora show
significant improvement over state-of-the-art methods.
|
Robust Classification by Pre-conditioned LASSO and Transductive
Diffusion Component Analysis | Modern machine learning-based recognition approaches require large-scale
datasets with large number of labelled training images. However, such datasets
are inherently difficult and costly to collect and annotate. Hence there is a
great and growing interest in automatic dataset collection methods that can
leverage the web. % which are collected % in a cheap, efficient and yet
unreliable way. Collecting datasets in this way, however, requires robust and
efficient ways for detecting and excluding outliers that are common and
prevalent. % Outliers are thus a % prominent treat of using these dataset. So
far, there have been a limited effort in machine learning community to directly
detect outliers for robust classification. Inspired by the recent work on
Pre-conditioned LASSO, this paper formulates the outlier detection task using
Pre-conditioned LASSO and employs \red{unsupervised} transductive diffusion
component analysis to both integrate the topological structure of the data
manifold, from labeled and unlabeled instances, and reduce the feature
dimensionality. Synthetic experiments as well as results on two real-world
classification tasks show that our framework can robustly detect the outliers
and improve classification.
|
Actor-Mimic: Deep Multitask and Transfer Reinforcement Learning | The ability to act in multiple environments and transfer previous knowledge
to new situations can be considered a critical aspect of any intelligent agent.
Towards this goal, we define a novel method of multitask and transfer learning
that enables an autonomous agent to learn how to behave in multiple tasks
simultaneously, and then generalize its knowledge to new domains. This method,
termed "Actor-Mimic", exploits the use of deep reinforcement learning and model
compression techniques to train a single policy network that learns how to act
in a set of distinct tasks by using the guidance of several expert teachers. We
then show that the representations learnt by the deep policy network are
capable of generalizing to new tasks with no prior expert guidance, speeding up
learning in novel environments. Although our method can in general be applied
to a wide range of problems, we use Atari games as a testing environment to
demonstrate these methods.
|
Online Batch Selection for Faster Training of Neural Networks | Deep neural networks are commonly trained using stochastic non-convex
optimization procedures, which are driven by gradient information estimated on
fractions (batches) of the dataset. While it is commonly accepted that batch
size is an important parameter for offline tuning, the benefits of online
selection of batches remain poorly understood. We investigate online batch
selection strategies for two state-of-the-art methods of stochastic
gradient-based optimization, AdaDelta and Adam. As the loss function to be
minimized for the whole dataset is an aggregation of loss functions of
individual datapoints, intuitively, datapoints with the greatest loss should be
considered (selected in a batch) more frequently. However, the limitations of
this intuition and the proper control of the selection pressure over time are
open questions. We propose a simple strategy where all datapoints are ranked
w.r.t. their latest known loss value and the probability to be selected decays
exponentially as a function of rank. Our experimental results on the MNIST
dataset suggest that selecting batches speeds up both AdaDelta and Adam by a
factor of about 5.
|
How much data is needed to train a medical image deep learning system to
achieve necessary high accuracy? | The use of Convolutional Neural Networks (CNN) in natural image
classification systems has produced very impressive results. Combined with the
inherent nature of medical images that make them ideal for deep-learning,
further application of such systems to medical image classification holds much
promise. However, the usefulness and potential impact of such a system can be
completely negated if it does not reach a target accuracy. In this paper, we
present a study on determining the optimum size of the training data set
necessary to achieve high classification accuracy with low variance in medical
image classification systems. The CNN was applied to classify axial Computed
Tomography (CT) images into six anatomical classes. We trained the CNN using
six different sizes of training data set (5, 10, 20, 50, 100, and 200) and then
tested the resulting system with a total of 6000 CT images. All images were
acquired from the Massachusetts General Hospital (MGH) Picture Archiving and
Communication System (PACS). Using this data, we employ the learning curve
approach to predict classification accuracy at a given training sample size.
Our research will present a general methodology for determining the training
data set size necessary to achieve a certain target classification accuracy
that can be easily applied to other problems within such systems.
|
Generating Sentences from a Continuous Space | The standard recurrent neural network language model (RNNLM) generates
sentences one word at a time and does not work from an explicit global sentence
representation. In this work, we introduce and study an RNN-based variational
autoencoder generative model that incorporates distributed latent
representations of entire sentences. This factorization allows it to explicitly
model holistic properties of sentences such as style, topic, and high-level
syntactic features. Samples from the prior over these sentence representations
remarkably produce diverse and well-formed sentences through simple
deterministic decoding. By examining paths through this latent space, we are
able to generate coherent novel sentences that interpolate between known
sentences. We present techniques for solving the difficult learning problem
presented by this model, demonstrate its effectiveness in imputing missing
words, explore many interesting properties of the model's latent sentence
space, and present negative results on the use of the model in language
modeling.
|
Structured Prediction Energy Networks | We introduce structured prediction energy networks (SPENs), a flexible
framework for structured prediction. A deep architecture is used to define an
energy function of candidate labels, and then predictions are produced by using
back-propagation to iteratively optimize the energy with respect to the labels.
This deep architecture captures dependencies between labels that would lead to
intractable graphical models, and performs structure learning by automatically
learning discriminative features of the structured output. One natural
application of our technique is multi-label classification, which traditionally
has required strict prior assumptions about the interactions between labels to
ensure tractable learning and prediction. We are able to apply SPENs to
multi-label problems with substantially larger label sets than previous
applications of structured prediction, while modeling high-order interactions
using minimal structural assumptions. Overall, deep learning provides
remarkable tools for learning features of the inputs to a prediction problem,
and this work extends these techniques to learning features of structured
outputs. Our experiments provide impressive performance on a variety of
benchmark multi-label classification tasks, demonstrate that our technique can
be used to provide interpretable structure learning, and illuminate fundamental
trade-offs between feed-forward and iterative structured prediction.
|
Learning Representations Using Complex-Valued Nets | Complex-valued neural networks (CVNNs) are an emerging field of research in
neural networks due to their potential representational properties for audio,
image, and physiological signals. It is common in signal processing to
transform sequences of real values to the complex domain via a set of complex
basis functions, such as the Fourier transform. We show how CVNNs can be used
to learn complex representations of real valued time-series data. We present
methods and results using a framework that can compose holomorphic and
non-holomorphic functions in a multi-layer network using a theoretical result
called the Wirtinger derivative. We test our methods on a representation
learning task for real-valued signals, recurrent complex-valued networks and
their real-valued counterparts. Our results show that recurrent complex-valued
networks can perform as well as their real-valued counterparts while learning
filters that are representative of the domain of the data.
|
FRIST - Flipping and Rotation Invariant Sparsifying Transform Learning
and Applications | Features based on sparse representation, especially using the synthesis
dictionary model, have been heavily exploited in signal processing and computer
vision. However, synthesis dictionary learning typically involves NP-hard
sparse coding and expensive learning steps. Recently, sparsifying transform
learning received interest for its cheap computation and its optimal updates in
the alternating algorithms. In this work, we develop a methodology for learning
Flipping and Rotation Invariant Sparsifying Transforms, dubbed FRIST, to better
represent natural images that contain textures with various geometrical
directions. The proposed alternating FRIST learning algorithm involves
efficient optimal updates. We provide a convergence guarantee, and demonstrate
the empirical convergence behavior of the proposed FRIST learning approach.
Preliminary experiments show the promising performance of FRIST learning for
sparse image representation, segmentation, denoising, robust inpainting, and
compressed sensing-based magnetic resonance image reconstruction.
|
Order-Embeddings of Images and Language | Hypernymy, textual entailment, and image captioning can be seen as special
cases of a single visual-semantic hierarchy over words, sentences, and images.
In this paper we advocate for explicitly modeling the partial order structure
of this hierarchy. Towards this goal, we introduce a general method for
learning ordered representations, and show how it can be applied to a variety
of tasks involving images and language. We show that the resulting
representations improve performance over current approaches for hypernym
prediction and image-caption retrieval.
|
Efficient inference in occlusion-aware generative models of images | We present a generative model of images based on layering, in which image
layers are individually generated, then composited from front to back. We are
thus able to factor the appearance of an image into the appearance of
individual objects within the image --- and additionally for each individual
object, we can factor content from pose. Unlike prior work on layered models,
we learn a shape prior for each object/layer, allowing the model to tease out
which object is in front by looking for a consistent shape, without needing
access to motion cues or any labeled data. We show that ordinary stochastic
gradient variational bayes (SGVB), which optimizes our fully differentiable
lower-bound on the log-likelihood, is sufficient to learn an interpretable
representation of images. Finally we present experiments demonstrating the
effectiveness of the model for inferring foreground and background objects in
images.
|
Dynamic Adaptive Network Intelligence | Accurate representational learning of both the explicit and implicit
relationships within data is critical to the ability of machines to perform
more complex and abstract reasoning tasks. We describe the efficient weakly
supervised learning of such inferences by our Dynamic Adaptive Network
Intelligence (DANI) model. We report state-of-the-art results for DANI over
question answering tasks in the bAbI dataset that have proved difficult for
contemporary approaches to learning representation (Weston et al., 2015).
|
Unsupervised Learning of Visual Structure using Predictive Generative
Networks | The ability to predict future states of the environment is a central pillar
of intelligence. At its core, effective prediction requires an internal model
of the world and an understanding of the rules by which the world changes.
Here, we explore the internal models developed by deep neural networks trained
using a loss based on predicting future frames in synthetic video sequences,
using a CNN-LSTM-deCNN framework. We first show that this architecture can
achieve excellent performance in visual sequence prediction tasks, including
state-of-the-art performance in a standard 'bouncing balls' dataset (Sutskever
et al., 2009). Using a weighted mean-squared error and adversarial loss
(Goodfellow et al., 2014), the same architecture successfully extrapolates
out-of-the-plane rotations of computer-generated faces. Furthermore, despite
being trained end-to-end to predict only pixel-level information, our
Predictive Generative Networks learn a representation of the latent structure
of the underlying three-dimensional objects themselves. Importantly, we find
that this representation is naturally tolerant to object transformations, and
generalizes well to new tasks, such as classification of static images. Similar
models trained solely with a reconstruction loss fail to generalize as
effectively. We argue that prediction can serve as a powerful unsupervised loss
for learning rich internal representations of high-level object features.
|
Manifold Regularized Deep Neural Networks using Adversarial Examples | Learning meaningful representations using deep neural networks involves
designing efficient training schemes and well-structured networks. Currently,
the method of stochastic gradient descent that has a momentum with dropout is
one of the most popular training protocols. Based on that, more advanced
methods (i.e., Maxout and Batch Normalization) have been proposed in recent
years, but most still suffer from performance degradation caused by small
perturbations, also known as adversarial examples. To address this issue, we
propose manifold regularized networks (MRnet) that utilize a novel training
objective function that minimizes the difference between multi-layer embedding
results of samples and those adversarial. Our experimental results demonstrated
that MRnet is more resilient to adversarial examples and helps us to generalize
representations on manifolds. Furthermore, combining MRnet and dropout allowed
us to achieve competitive classification performances for three well-known
benchmarks: MNIST, CIFAR-10, and SVHN.
|
Iterative Refinement of the Approximate Posterior for Directed Belief
Networks | Variational methods that rely on a recognition network to approximate the
posterior of directed graphical models offer better inference and learning than
previous methods. Recent advances that exploit the capacity and flexibility in
this approach have expanded what kinds of models can be trained. However, as a
proposal for the posterior, the capacity of the recognition network is limited,
which can constrain the representational power of the generative model and
increase the variance of Monte Carlo estimates. To address these issues, we
introduce an iterative refinement procedure for improving the approximate
posterior of the recognition network and show that training with the refined
posterior is competitive with state-of-the-art methods. The advantages of
refinement are further evident in an increased effective sample size, which
implies a lower variance of gradient estimates.
|
A Unified Gradient Regularization Family for Adversarial Examples | Adversarial examples are augmented data points generated by imperceptible
perturbation of input samples. They have recently drawn much attention with the
machine learning and data mining community. Being difficult to distinguish from
real examples, such adversarial examples could change the prediction of many of
the best learning models including the state-of-the-art deep learning models.
Recent attempts have been made to build robust models that take into account
adversarial examples. However, these methods can either lead to performance
drops or lack mathematical motivations. In this paper, we propose a unified
framework to build robust machine learning models against adversarial examples.
More specifically, using the unified framework, we develop a family of gradient
regularization methods that effectively penalize the gradient of loss function
w.r.t. inputs. Our proposed framework is appealing in that it offers a unified
view to deal with adversarial examples. It incorporates another
recently-proposed perturbation based approach as a special case. In addition,
we present some visual effects that reveals semantic meaning in those
perturbations, and thus support our regularization method and provide another
explanation for generalizability of adversarial examples. By applying this
technique to Maxout networks, we conduct a series of experiments and achieve
encouraging results on two benchmark datasets. In particular,we attain the best
accuracy on MNIST data (without data augmentation) and competitive performance
on CIFAR-10 data.
|
sense2vec - A Fast and Accurate Method for Word Sense Disambiguation In
Neural Word Embeddings | Neural word representations have proven useful in Natural Language Processing
(NLP) tasks due to their ability to efficiently model complex semantic and
syntactic word relationships. However, most techniques model only one
representation per word, despite the fact that a single word can have multiple
meanings or "senses". Some techniques model words by using multiple vectors
that are clustered based on context. However, recent neural approaches rarely
focus on the application to a consuming NLP algorithm. Furthermore, the
training process of recent word-sense models is expensive relative to
single-sense embedding processes. This paper presents a novel approach which
addresses these concerns by modeling multiple embeddings for each word based on
supervised disambiguation, which provides a fast and accurate way for a
consuming NLP model to select a sense-disambiguated embedding. We demonstrate
that these embeddings can disambiguate both contrastive senses such as nominal
and verbal senses as well as nuanced senses such as sarcasm. We further
evaluate Part-of-Speech disambiguated embeddings on neural dependency parsing,
yielding a greater than 8% average error reduction in unlabeled attachment
scores across 6 languages.
|
Unsupervised and Semi-supervised Learning with Categorical Generative
Adversarial Networks | In this paper we present a method for learning a discriminative classifier
from unlabeled or partially labeled data. Our approach is based on an objective
function that trades-off mutual information between observed examples and their
predicted categorical class distribution, against robustness of the classifier
to an adversarial generative model. The resulting algorithm can either be
interpreted as a natural generalization of the generative adversarial networks
(GAN) framework or as an extension of the regularized information maximization
(RIM) framework to robust classification against an optimal adversary. We
empirically evaluate our method - which we dub categorical generative
adversarial networks (or CatGAN) - on synthetic data as well as on challenging
image classification tasks, demonstrating the robustness of the learned
classifiers. We further qualitatively assess the fidelity of samples generated
by the adversarial generator that is learned alongside the discriminative
classifier, and identify links between the CatGAN objective and discriminative
clustering algorithms (such as RIM).
|
Order Matters: Sequence to sequence for sets | Sequences have become first class citizens in supervised learning thanks to
the resurgence of recurrent neural networks. Many complex tasks that require
mapping from or to a sequence of observations can now be formulated with the
sequence-to-sequence (seq2seq) framework which employs the chain rule to
efficiently represent the joint probability of sequences. In many cases,
however, variable sized inputs and/or outputs might not be naturally expressed
as sequences. For instance, it is not clear how to input a set of numbers into
a model where the task is to sort them; similarly, we do not know how to
organize outputs when they correspond to random variables and the task is to
model their unknown joint probability. In this paper, we first show using
various examples that the order in which we organize input and/or output data
matters significantly when learning an underlying model. We then discuss an
extension of the seq2seq framework that goes beyond sequences and handles input
sets in a principled way. In addition, we propose a loss which, by searching
over possible orders during training, deals with the lack of structure of
output sets. We show empirical evidence of our claims regarding ordering, and
on the modifications to the seq2seq framework on benchmark language modeling
and parsing tasks, as well as two artificial tasks -- sorting numbers and
estimating the joint probability of unknown graphical models.
|
Neural Random-Access Machines | In this paper, we propose and investigate a new neural network architecture
called Neural Random Access Machine. It can manipulate and dereference pointers
to an external variable-size random-access memory. The model is trained from
pure input-output examples using backpropagation.
We evaluate the new model on a number of simple algorithmic tasks whose
solutions require pointer manipulation and dereferencing. Our results show that
the proposed model can learn to solve algorithmic tasks of such type and is
capable of operating on simple data structures like linked-lists and binary
trees. For easier tasks, the learned solutions generalize to sequences of
arbitrary length. Moreover, memory access during inference can be done in a
constant time under some assumptions.
|
Fixed Point Quantization of Deep Convolutional Networks | In recent years increasingly complex architectures for deep convolution
networks (DCNs) have been proposed to boost the performance on image
recognition tasks. However, the gains in performance have come at a cost of
substantial increase in computation and model storage resources. Fixed point
implementation of DCNs has the potential to alleviate some of these
complexities and facilitate potential deployment on embedded hardware. In this
paper, we propose a quantizer design for fixed point implementation of DCNs. We
formulate and solve an optimization problem to identify optimal fixed point
bit-width allocation across DCN layers. Our experiments show that in comparison
to equal bit-width settings, the fixed point DCNs with optimized bit width
allocation offer >20% reduction in the model size without any loss in accuracy
on CIFAR-10 benchmark. We also demonstrate that fine-tuning can further enhance
the accuracy of fixed point DCNs beyond that of the original floating point
model. In doing so, we report a new state-of-the-art fixed point performance of
6.78% error-rate on CIFAR-10 benchmark.
|
Geodesics of learned representations | We develop a new method for visualizing and refining the invariances of
learned representations. Specifically, we test for a general form of
invariance, linearization, in which the action of a transformation is confined
to a low-dimensional subspace. Given two reference images (typically, differing
by some transformation), we synthesize a sequence of images lying on a path
between them that is of minimal length in the space of the representation (a
"representational geodesic"). If the transformation relating the two reference
images is linearized by the representation, this sequence should follow the
gradual evolution of this transformation. We use this method to assess the
invariance properties of a state-of-the-art image classification network and
find that geodesics generated for image pairs differing by translation,
rotation, and dilation do not evolve according to their associated
transformations. Our method also suggests a remedy for these failures, and
following this prescription, we show that the modified representation is able
to linearize a variety of geometric image transformations.
|
Multilingual Relation Extraction using Compositional Universal Schema | Universal schema builds a knowledge base (KB) of entities and relations by
jointly embedding all relation types from input KBs as well as textual patterns
expressing relations from raw text. In most previous applications of universal
schema, each textual pattern is represented as a single embedding, preventing
generalization to unseen patterns. Recent work employs a neural network to
capture patterns' compositional semantics, providing generalization to all
possible input text. In response, this paper introduces significant further
improvements to the coverage and flexibility of universal schema relation
extraction: predictions for entities unseen in training and multilingual
transfer learning to domains with no annotation. We evaluate our model through
extensive experiments on the English and Spanish TAC KBP benchmark,
outperforming the top system from TAC 2013 slot-filling using no handwritten
patterns or additional annotation. We also consider a multilingual setting in
which English training data entities overlap with the seed KB, but Spanish text
does not. Despite having no annotation for Spanish data, we train an accurate
predictor, with additional improvements obtained by tying word embeddings
across languages. Furthermore, we find that multilingual training improves
English relation extraction accuracy. Our approach is thus suited to
broad-coverage automated knowledge base construction in a variety of languages
and domains.
|
Compressing Word Embeddings | Recent methods for learning vector space representations of words have
succeeded in capturing fine-grained semantic and syntactic regularities using
vector arithmetic. However, these vector space representations (created through
large-scale text analysis) are typically stored verbatim, since their internal
structure is opaque. Using word-analogy tests to monitor the level of detail
stored in compressed re-representations of the same vector space, the
trade-offs between the reduction in memory usage and expressiveness are
investigated. A simple scheme is outlined that can reduce the memory footprint
of a state-of-the-art embedding by a factor of 10, with only minimal impact on
performance. Then, using the same `bit budget', a binary (approximate)
factorisation of the same space is also explored, with the aim of creating an
equivalent representation with better interpretability.
|
Denoising Criterion for Variational Auto-Encoding Framework | Denoising autoencoders (DAE) are trained to reconstruct their clean inputs
with noise injected at the input level, while variational autoencoders (VAE)
are trained with noise injected in their stochastic hidden layer, with a
regularizer that encourages this noise injection. In this paper, we show that
injecting noise both in input and in the stochastic hidden layer can be
advantageous and we propose a modified variational lower bound as an improved
objective function in this setup. When input is corrupted, then the standard
VAE lower bound involves marginalizing the encoder conditional distribution
over the input noise, which makes the training criterion intractable. Instead,
we propose a modified training criterion which corresponds to a tractable bound
when input is corrupted. Experimentally, we find that the proposed denoising
variational autoencoder (DVAE) yields better average log-likelihood than the
VAE and the importance weighted autoencoder on the MNIST and Frey Face
datasets.
|
Recurrent Models for Auditory Attention in Multi-Microphone Distance
Speech Recognition | Integration of multiple microphone data is one of the key ways to achieve
robust speech recognition in noisy environments or when the speaker is located
at some distance from the input device. Signal processing techniques such as
beamforming are widely used to extract a speech signal of interest from
background noise. These techniques, however, are highly dependent on prior
spatial information about the microphones and the environment in which the
system is being used. In this work, we present a neural attention network that
directly combines multi-channel audio to generate phonetic states without
requiring any prior knowledge of the microphone layout or any explicit signal
preprocessing for speech enhancement. We embed an attention mechanism within a
Recurrent Neural Network (RNN) based acoustic model to automatically tune its
attention to a more reliable input source. Unlike traditional multi-channel
preprocessing, our system can be optimized towards the desired output in one
step. Although attention-based models have recently achieved impressive results
on sequence-to-sequence learning, no attention mechanisms have previously been
applied to learn potentially asynchronous and non-stationary multiple inputs.
We evaluate our neural attention model on the CHiME-3 challenge task, and show
that the model achieves comparable performance to beamforming using a purely
data-driven method.
|
Learning to Generate Images with Perceptual Similarity Metrics | Deep networks are increasingly being applied to problems involving image
synthesis, e.g., generating images from textual descriptions and reconstructing
an input image from a compact representation. Supervised training of
image-synthesis networks typically uses a pixel-wise loss (PL) to indicate the
mismatch between a generated image and its corresponding target image. We
propose instead to use a loss function that is better calibrated to human
perceptual judgments of image quality: the multiscale structural-similarity
score (MS-SSIM). Because MS-SSIM is differentiable, it is easily incorporated
into gradient-descent learning. We compare the consequences of using MS-SSIM
versus PL loss on training deterministic and stochastic autoencoders. For three
different architectures, we collected human judgments of the quality of image
reconstructions. Observers reliably prefer images synthesized by
MS-SSIM-optimized models over those synthesized by PL-optimized models, for two
distinct PL measures ($\ell_1$ and $\ell_2$ distances). We also explore the
effect of training objective on image encoding and analyze conditions under
which perceptually-optimized representations yield better performance on image
classification. Finally, we demonstrate the superiority of
perceptually-optimized networks for super-resolution imaging. Just as computer
vision has advanced through the use of convolutional architectures that mimic
the structure of the mammalian visual system, we argue that significant
additional advances can be made in modeling images through the use of training
objectives that are well aligned to characteristics of human perception.
|
Better Computer Go Player with Neural Network and Long-term Prediction | Competing with top human players in the ancient game of Go has been a
long-term goal of artificial intelligence. Go's high branching factor makes
traditional search techniques ineffective, even on leading-edge hardware, and
Go's evaluation function could change drastically with one stone change. Recent
works [Maddison et al. (2015); Clark & Storkey (2015)] show that search is not
strictly necessary for machine Go players. A pure pattern-matching approach,
based on a Deep Convolutional Neural Network (DCNN) that predicts the next
move, can perform as well as Monte Carlo Tree Search (MCTS)-based open source
Go engines such as Pachi [Baudis & Gailly (2012)] if its search budget is
limited. We extend this idea in our bot named darkforest, which relies on a
DCNN designed for long-term predictions. Darkforest substantially improves the
win rate for pattern-matching approaches against MCTS-based approaches, even
with looser search budgets. Against human players, the newest versions,
darkfores2, achieve a stable 3d level on KGS Go Server as a ranked bot, a
substantial improvement upon the estimated 4k-5k ranks for DCNN reported in
Clark & Storkey (2015) based on games against other machine players. Adding
MCTS to darkfores2 creates a much stronger player named darkfmcts3: with 5000
rollouts, it beats Pachi with 10k rollouts in all 250 games; with 75k rollouts
it achieves a stable 5d level in KGS server, on par with state-of-the-art Go
AIs (e.g., Zen, DolBaram, CrazyStone) except for AlphaGo [Silver et al.
(2016)]; with 110k rollouts, it won the 3rd place in January KGS Go Tournament.
|
Training Deep Neural Networks via Direct Loss Minimization | Supervised training of deep neural nets typically relies on minimizing
cross-entropy. However, in many domains, we are interested in performing well
on metrics specific to the application. In this paper we propose a direct loss
minimization approach to train deep neural networks, which provably minimizes
the application-specific loss function. This is often non-trivial, since these
functions are neither smooth nor decomposable and thus are not amenable to
optimization with standard gradient-based methods. We demonstrate the
effectiveness of our approach in the context of maximizing average precision
for ranking problems. Towards this goal, we develop a novel dynamic programming
algorithm that can efficiently compute the weight updates. Our approach proves
superior to a variety of baselines in the context of action classification and
object detection, especially in the presence of label noise.
|
QBDC: Query by dropout committee for training deep supervised
architecture | While the current trend is to increase the depth of neural networks to
increase their performance, the size of their training database has to grow
accordingly. We notice an emergence of tremendous databases, although providing
labels to build a training set still remains a very expensive task. We tackle
the problem of selecting the samples to be labelled in an online fashion. In
this paper, we present an active learning strategy based on query by committee
and dropout technique to train a Convolutional Neural Network (CNN). We derive
a commmittee of partial CNNs resulting from batchwise dropout runs on the
initial CNN. We evaluate our active learning strategy for CNN on MNIST
benchmark, showing in particular that selecting less than 30 % from the
annotated database is enough to get similar error rate as using the full
training set on MNIST. We also studied the robustness of our method against
adversarial examples.
|
Fast Parallel SAME Gibbs Sampling on General Discrete Bayesian Networks | A fundamental task in machine learning and related fields is to perform
inference on Bayesian networks. Since exact inference takes exponential time in
general, a variety of approximate methods are used. Gibbs sampling is one of
the most accurate approaches and provides unbiased samples from the posterior
but it has historically been too expensive for large models. In this paper, we
present an optimized, parallel Gibbs sampler augmented with state replication
(SAME or State Augmented Marginal Estimation) to decrease convergence time. We
find that SAME can improve the quality of parameter estimates while
accelerating convergence. Experiments on both synthetic and real data show that
our Gibbs sampler is substantially faster than the state of the art sampler,
JAGS, without sacrificing accuracy. Our ultimate objective is to introduce the
Gibbs sampler to researchers in many fields to expand their range of feasible
inference problems.
|
Binding via Reconstruction Clustering | Disentangled distributed representations of data are desirable for machine
learning, since they are more expressive and can generalize from fewer
examples. However, for complex data, the distributed representations of
multiple objects present in the same input can interfere and lead to
ambiguities, which is commonly referred to as the binding problem. We argue for
the importance of the binding problem to the field of representation learning,
and develop a probabilistic framework that explicitly models inputs as a
composition of multiple objects. We propose an unsupervised algorithm that uses
denoising autoencoders to dynamically bind features together in multi-object
inputs through an Expectation-Maximization-like clustering process. The
effectiveness of this method is demonstrated on artificially generated datasets
of binary images, showing that it can even generalize to bind together new
objects never seen by the autoencoder during training.
|
Canonical Autocorrelation Analysis | We present an extension of sparse Canonical Correlation Analysis (CCA)
designed for finding multiple-to-multiple linear correlations within a single
set of variables. Unlike CCA, which finds correlations between two sets of data
where the rows are matched exactly but the columns represent separate sets of
variables, the method proposed here, Canonical Autocorrelation Analysis (CAA),
finds multivariate correlations within just one set of variables. This can be
useful when we look for hidden parsimonious structures in data, each involving
only a small subset of all features. In addition, the discovered correlations
are highly interpretable as they are formed by pairs of sparse linear
combinations of the original features. We show how CAA can be of use as a tool
for anomaly detection when the expected structure of correlations is not
followed by anomalous data. We illustrate the utility of CAA in two application
domains where single-class and unsupervised learning of correlation structures
are particularly relevant: breast cancer diagnosis and radiation threat
detection. When applied to the Wisconsin Breast Cancer data, single-class CAA
is competitive with supervised methods used in literature. On the radiation
threat detection task, unsupervised CAA performs significantly better than an
unsupervised alternative prevalent in the domain, while providing valuable
additional insights for threat analysis.
|
Skip-Thought Memory Networks | Question Answering (QA) is fundamental to natural language processing in that
most nlp problems can be phrased as QA (Kumar et al., 2015). Current weakly
supervised memory network models that have been proposed so far struggle at
answering questions that involve relations among multiple entities (such as
facebook's bAbi qa5-three-arg-relations in (Weston et al., 2015)). To address
this problem of learning multi-argument multi-hop semantic relations for the
purpose of QA, we propose a method that combines the jointly learned long-term
read-write memory and attentive inference components of end-to-end memory
networks (MemN2N) (Sukhbaatar et al., 2015) with distributed sentence vector
representations encoded by a Skip-Thought model (Kiros et al., 2015). This
choice to append Skip-Thought Vectors to the existing MemN2N framework is
motivated by the fact that Skip-Thought Vectors have been shown to accurately
model multi-argument semantic relations (Kiros et al., 2015).
|
Deep Manifold Traversal: Changing Labels with Convolutional Features | Many tasks in computer vision can be cast as a "label changing" problem,
where the goal is to make a semantic change to the appearance of an image or
some subject in an image in order to alter the class membership. Although
successful task-specific methods have been developed for some label changing
applications, to date no general purpose method exists. Motivated by this we
propose deep manifold traversal, a method that addresses the problem in its
most general form: it first approximates the manifold of natural images then
morphs a test image along a traversal path away from a source class and towards
a target class while staying near the manifold throughout. The resulting
algorithm is surprisingly effective and versatile. It is completely data
driven, requiring only an example set of images from the desired source and
target domains. We demonstrate deep manifold traversal on highly diverse label
changing tasks: changing an individual's appearance (age and hair color),
changing the season of an outdoor image, and transforming a city skyline
towards nighttime.
|
All you need is a good init | Layer-sequential unit-variance (LSUV) initialization - a simple method for
weight initialization for deep net learning - is proposed. The method consists
of the two steps. First, pre-initialize weights of each convolution or
inner-product layer with orthonormal matrices. Second, proceed from the first
to the final layer, normalizing the variance of the output of each layer to be
equal to one.
Experiment with different activation functions (maxout, ReLU-family, tanh)
show that the proposed initialization leads to learning of very deep nets that
(i) produces networks with test accuracy better or equal to standard methods
and (ii) is at least as fast as the complex schemes proposed specifically for
very deep nets such as FitNets (Romero et al. (2015)) and Highway (Srivastava
et al. (2015)).
Performance is evaluated on GoogLeNet, CaffeNet, FitNets and Residual nets
and the state-of-the-art, or very close to it, is achieved on the MNIST,
CIFAR-10/100 and ImageNet datasets.
|
An Information Retrieval Approach to Finding Dependent Subspaces of
Multiple Views | Finding relationships between multiple views of data is essential both for
exploratory analysis and as pre-processing for predictive tasks. A prominent
approach is to apply variants of Canonical Correlation Analysis (CCA), a
classical method seeking correlated components between views. The basic CCA is
restricted to maximizing a simple dependency criterion, correlation, measured
directly between data coordinates. We introduce a new method that finds
dependent subspaces of views directly optimized for the data analysis task of
\textit{neighbor retrieval between multiple views}. We optimize mappings for
each view such as linear transformations to maximize cross-view similarity
between neighborhoods of data samples. The criterion arises directly from the
well-defined retrieval task, detects nonlinear and local similarities, is able
to measure dependency of data relationships rather than only individual data
coordinates, and is related to well understood measures of information
retrieval quality. In experiments we show the proposed method outperforms
alternatives in preserving cross-view neighborhood similarities, and yields
insights into local dependencies between multiple views.
|
First Step toward Model-Free, Anonymous Object Tracking with Recurrent
Neural Networks | In this paper, we propose and study a novel visual object tracking approach
based on convolutional networks and recurrent networks. The proposed approach
is distinct from the existing approaches to visual object tracking, such as
filtering-based ones and tracking-by-detection ones, in the sense that the
tracking system is explicitly trained off-line to track anonymous objects in a
noisy environment. The proposed visual tracking model is end-to-end trainable,
minimizing any adversarial effect from mismatches in object representation and
between the true underlying dynamics and learning dynamics. We empirically show
that the proposed tracking approach works well in various scenarios by
generating artificial video sequences with varying conditions; the number of
objects, amount of noise and the match between the training shapes and test
shapes.
|
A Controller-Recognizer Framework: How necessary is recognition for
control? | Recently there has been growing interest in building active visual object
recognizers, as opposed to the usual passive recognizers which classifies a
given static image into a predefined set of object categories. In this paper we
propose to generalize these recently proposed end-to-end active visual
recognizers into a controller-recognizer framework. A model in the
controller-recognizer framework consists of a controller, which interfaces with
an external manipulator, and a recognizer which classifies the visual input
adjusted by the manipulator. We describe two most recently proposed
controller-recognizer models: recurrent attention model and spatial transformer
network as representative examples of controller-recognizer models. Based on
this description we observe that most existing end-to-end
controller-recognizers tightly, or completely, couple a controller and
recognizer. We ask a question whether this tight coupling is necessary, and try
to answer this empirically by building a controller-recognizer model with a
decoupled controller and recognizer. Our experiments revealed that it is not
always necessary to tightly couple them and that by decoupling a controller and
recognizer, there is a possibility of building a generic controller that is
pretrained and works together with any subsequent recognizer.
|
Patterns for Learning with Side Information | Supervised, semi-supervised, and unsupervised learning estimate a function
given input/output samples. Generalization of the learned function to unseen
data can be improved by incorporating side information into learning. Side
information are data that are neither from the input space nor from the output
space of the function, but include useful information for learning it. In this
paper we show that learning with side information subsumes a variety of related
approaches, e.g. multi-task learning, multi-view learning and learning using
privileged information. Our main contributions are (i) a new perspective that
connects these previously isolated approaches, (ii) insights about how these
methods incorporate different types of prior knowledge, and hence implement
different patterns, (iii) facilitating the application of these methods in
novel tasks, as well as (iv) a systematic experimental evaluation of these
patterns in two supervised learning tasks.
|
Deconstructing the Ladder Network Architecture | The Manual labeling of data is and will remain a costly endeavor. For this
reason, semi-supervised learning remains a topic of practical importance. The
recently proposed Ladder Network is one such approach that has proven to be
very successful. In addition to the supervised objective, the Ladder Network
also adds an unsupervised objective corresponding to the reconstruction costs
of a stack of denoising autoencoders. Although the empirical results are
impressive, the Ladder Network has many components intertwined, whose
contributions are not obvious in such a complex architecture. In order to help
elucidate and disentangle the different ingredients in the Ladder Network
recipe, this paper presents an extensive experimental investigation of variants
of the Ladder Network in which we replace or remove individual components to
gain more insight into their relative importance. We find that all of the
components are necessary for achieving optimal performance, but they do not
contribute equally. For semi-supervised tasks, we conclude that the most
important contribution is made by the lateral connection, followed by the
application of noise, and finally the choice of what we refer to as the
`combinator function' in the decoder path. We also find that as the number of
labeled training examples increases, the lateral connections and reconstruction
criterion become less important, with most of the improvement in generalization
being due to the injection of noise in each layer. Furthermore, we present a
new type of combinator function that outperforms the original design in both
fully- and semi-supervised tasks, reducing record test error rates on
Permutation-Invariant MNIST to 0.57% for the supervised setting, and to 0.97%
and 1.0% for semi-supervised settings with 1000 and 100 labeled examples
respectively.
|
Delving Deeper into Convolutional Networks for Learning Video
Representations | We propose an approach to learn spatio-temporal features in videos from
intermediate visual representations we call "percepts" using
Gated-Recurrent-Unit Recurrent Networks (GRUs).Our method relies on percepts
that are extracted from all level of a deep convolutional network trained on
the large ImageNet dataset. While high-level percepts contain highly
discriminative information, they tend to have a low-spatial resolution.
Low-level percepts, on the other hand, preserve a higher spatial resolution
from which we can model finer motion patterns. Using low-level percepts can
leads to high-dimensionality video representations. To mitigate this effect and
control the model number of parameters, we introduce a variant of the GRU model
that leverages the convolution operations to enforce sparse connectivity of the
model units and share parameters across the input spatial locations.
We empirically validate our approach on both Human Action Recognition and
Video Captioning tasks. In particular, we achieve results equivalent to
state-of-art on the YouTube2Text dataset using a simpler text-decoder model and
without extra 3D CNN features.
|
Blending LSTMs into CNNs | We consider whether deep convolutional networks (CNNs) can represent decision
functions with similar accuracy as recurrent networks such as LSTMs. First, we
show that a deep CNN with an architecture inspired by the models recently
introduced in image recognition can yield better accuracy than previous
convolutional and LSTM networks on the standard 309h Switchboard automatic
speech recognition task. Then we show that even more accurate CNNs can be
trained under the guidance of LSTMs using a variant of model compression, which
we call model blending because the teacher and student models are similar in
complexity but different in inductive bias. Blending further improves the
accuracy of our CNN, yielding a computationally efficient model of accuracy
higher than any of the other individual models. Examining the effect of "dark
knowledge" in this model compression task, we find that less than 1% of the
highest probability labels are needed for accurate model compression.
|
Unsupervised Representation Learning with Deep Convolutional Generative
Adversarial Networks | In recent years, supervised learning with convolutional networks (CNNs) has
seen huge adoption in computer vision applications. Comparatively, unsupervised
learning with CNNs has received less attention. In this work we hope to help
bridge the gap between the success of CNNs for supervised learning and
unsupervised learning. We introduce a class of CNNs called deep convolutional
generative adversarial networks (DCGANs), that have certain architectural
constraints, and demonstrate that they are a strong candidate for unsupervised
learning. Training on various image datasets, we show convincing evidence that
our deep convolutional adversarial pair learns a hierarchy of representations
from object parts to scenes in both the generator and discriminator.
Additionally, we use the learned features for novel tasks - demonstrating their
applicability as general image representations.
|
Comparative Study of Deep Learning Software Frameworks | Deep learning methods have resulted in significant performance improvements
in several application domains and as such several software frameworks have
been developed to facilitate their implementation. This paper presents a
comparative study of five deep learning frameworks, namely Caffe, Neon,
TensorFlow, Theano, and Torch, on three aspects: extensibility, hardware
utilization, and speed. The study is performed on several types of deep
learning architectures and we evaluate the performance of the above frameworks
when employed on a single machine for both (multi-threaded) CPU and GPU (Nvidia
Titan X) settings. The speed performance metrics used here include the gradient
computation time, which is important during the training phase of deep
networks, and the forward time, which is important from the deployment
perspective of trained networks. For convolutional networks, we also report how
each of these frameworks support various convolutional algorithms and their
corresponding performance. From our experiments, we observe that Theano and
Torch are the most easily extensible frameworks. We observe that Torch is best
suited for any deep architecture on CPU, followed by Theano. It also achieves
the best performance on the GPU for large convolutional and fully connected
networks, followed closely by Neon. Theano achieves the best performance on GPU
for training and deployment of LSTM networks. Caffe is the easiest for
evaluating the performance of standard deep architectures. Finally, TensorFlow
is a very flexible framework, similar to Theano, but its performance is
currently not competitive compared to the other studied frameworks.
|
A convnet for non-maximum suppression | Non-maximum suppression (NMS) is used in virtually all state-of-the-art
object detection pipelines. While essential object detection ingredients such
as features, classifiers, and proposal methods have been extensively researched
surprisingly little work has aimed to systematically address NMS. The de-facto
standard for NMS is based on greedy clustering with a fixed distance threshold,
which forces to trade-off recall versus precision. We propose a convnet
designed to perform NMS of a given set of detections. We report experiments on
a synthetic setup, and results on crowded pedestrian detection scenes. Our
approach overcomes the intrinsic limitations of greedy NMS, obtaining better
recall and precision.
|
Towards Principled Unsupervised Learning | General unsupervised learning is a long-standing conceptual problem in
machine learning. Supervised learning is successful because it can be solved by
the minimization of the training error cost function. Unsupervised learning is
not as successful, because the unsupervised objective may be unrelated to the
supervised task of interest. For an example, density modelling and
reconstruction have often been used for unsupervised learning, but they did not
produced the sought-after performance gains, because they have no knowledge of
the supervised tasks.
In this paper, we present an unsupervised cost function which we name the
Output Distribution Matching (ODM) cost, which measures a divergence between
the distribution of predictions and distributions of labels. The ODM cost is
appealing because it is consistent with the supervised cost in the following
sense: a perfect supervised classifier is also perfect according to the ODM
cost. Therefore, by aggressively optimizing the ODM cost, we are almost
guaranteed to improve our supervised performance whenever the space of possible
predictions is exponentially large.
We demonstrate that the ODM cost works well on number of small and
semi-artificial datasets using no (or almost no) labelled training cases.
Finally, we show that the ODM cost can be used for one-shot domain adaptation,
which allows the model to classify inputs that differ from the input
distribution in significant ways without the need for prior exposure to the new
domain.
|
Fast Metric Learning For Deep Neural Networks | Similarity metrics are a core component of many information retrieval and
machine learning systems. In this work we propose a method capable of learning
a similarity metric from data equipped with a binary relation. By considering
only the similarity constraints, and initially ignoring the features, we are
able to learn target vectors for each instance using one of several
appropriately designed loss functions. A regression model can then be
constructed that maps novel feature vectors to the same target vector space,
resulting in a feature extractor that computes vectors for which a predefined
metric is a meaningful measure of similarity. We present results on both
multiclass and multi-label classification datasets that demonstrate
considerably faster convergence, as well as higher accuracy on the majority of
the intrinsic evaluation tasks and all extrinsic evaluation tasks.
|
Neural Network Matrix Factorization | Data often comes in the form of an array or matrix. Matrix factorization
techniques attempt to recover missing or corrupted entries by assuming that the
matrix can be written as the product of two low-rank matrices. In other words,
matrix factorization approximates the entries of the matrix by a simple, fixed
function---namely, the inner product---acting on the latent feature vectors for
the corresponding row and column. Here we consider replacing the inner product
by an arbitrary function that we learn from the data at the same time as we
learn the latent feature vectors. In particular, we replace the inner product
by a multi-layer feed-forward neural network, and learn by alternating between
optimizing the network for fixed latent features, and optimizing the latent
features for a fixed network. The resulting approach---which we call neural
network matrix factorization or NNMF, for short---dominates standard low-rank
techniques on a suite of benchmark but is dominated by some recent proposals
that take advantage of the graph features. Given the vast range of
architectures, activation functions, regularizers, and optimization techniques
that could be used within the NNMF framework, it seems likely the true
potential of the approach has yet to be reached.
|
Universal halting times in optimization and machine learning | The authors present empirical distributions for the halting time (measured by
the number of iterations to reach a given accuracy) of optimization algorithms
applied to two random systems: spin glasses and deep learning. Given an
algorithm, which we take to be both the optimization routine and the form of
the random landscape, the fluctuations of the halting time follow a
distribution that, after centering and scaling, remains unchanged even when the
distribution on the landscape is changed. We observe two qualitative classes: A
Gumbel-like distribution that appears in Google searches, human decision times,
the QR eigenvalue algorithm and spin glasses, and a Gaussian-like distribution
that appears in conjugate gradient method, deep network with MNIST input data
and deep network with random input data. This empirical evidence suggests
presence of a class of distributions for which the halting time is independent
of the underlying distribution under some conditions.
|
Learning Representations from EEG with Deep Recurrent-Convolutional
Neural Networks | One of the challenges in modeling cognitive events from electroencephalogram
(EEG) data is finding representations that are invariant to inter- and
intra-subject differences, as well as to inherent noise associated with such
data. Herein, we propose a novel approach for learning such representations
from multi-channel EEG time-series, and demonstrate its advantages in the
context of mental load classification task. First, we transform EEG activities
into a sequence of topology-preserving multi-spectral images, as opposed to
standard EEG analysis techniques that ignore such spatial information. Next, we
train a deep recurrent-convolutional network inspired by state-of-the-art video
classification to learn robust representations from the sequence of images. The
proposed approach is designed to preserve the spatial, spectral, and temporal
structure of EEG which leads to finding features that are less sensitive to
variations and distortions within each dimension. Empirical evaluation on the
cognitive load classification task demonstrated significant improvements in
classification accuracy over current state-of-the-art approaches in this field.
|
Learning to decompose for object detection and instance segmentation | Although deep convolutional neural networks(CNNs) have achieved remarkable
results on object detection and segmentation, pre- and post-processing steps
such as region proposals and non-maximum suppression(NMS), have been required.
These steps result in high computational complexity and sensitivity to
hyperparameters, e.g. thresholds for NMS. In this work, we propose a novel
end-to-end trainable deep neural network architecture, which consists of
convolutional and recurrent layers, that generates the correct number of object
instances and their bounding boxes (or segmentation masks) given an image,
using only a single network evaluation without any pre- or post-processing
steps. We have tested on detecting digits in multi-digit images synthesized
using MNIST, automatically segmenting digits in these images, and detecting
cars in the KITTI benchmark dataset. The proposed approach outperforms a strong
CNN baseline on the synthesized digits datasets and shows promising results on
KITTI car detection.
|
Deep Metric Learning via Lifted Structured Feature Embedding | Learning the distance metric between pairs of examples is of great importance
for learning and visual recognition. With the remarkable success from the state
of the art convolutional neural networks, recent works have shown promising
results on discriminatively training the networks to learn semantic feature
embeddings where similar examples are mapped close to each other and dissimilar
examples are mapped farther apart. In this paper, we describe an algorithm for
taking full advantage of the training batches in the neural network training by
lifting the vector of pairwise distances within the batch to the matrix of
pairwise distances. This step enables the algorithm to learn the state of the
art feature embedding by optimizing a novel structured prediction objective on
the lifted problem. Additionally, we collected Online Products dataset: 120k
images of 23k classes of online products for metric learning. Our experiments
on the CUB-200-2011, CARS196, and Online Products datasets demonstrate
significant improvement over existing deep feature embedding methods on all
experimented embedding sizes with the GoogLeNet network.
|
Variational Auto-encoded Deep Gaussian Processes | We develop a scalable deep non-parametric generative model by augmenting deep
Gaussian processes with a recognition model. Inference is performed in a novel
scalable variational framework where the variational posterior distributions
are reparametrized through a multilayer perceptron. The key aspect of this
reformulation is that it prevents the proliferation of variational parameters
which otherwise grow linearly in proportion to the sample size. We derive a new
formulation of the variational lower bound that allows us to distribute most of
the computation in a way that enables to handle datasets of the size of
mainstream deep learning tasks. We show the efficacy of the method on a variety
of challenges including deep unsupervised learning and deep Bayesian
optimization.
|
Task Loss Estimation for Sequence Prediction | Often, the performance on a supervised machine learning task is evaluated
with a emph{task loss} function that cannot be optimized directly. Examples of
such loss functions include the classification error, the edit distance and the
BLEU score. A common workaround for this problem is to instead optimize a
emph{surrogate loss} function, such as for instance cross-entropy or hinge
loss. In order for this remedy to be effective, it is important to ensure that
minimization of the surrogate loss results in minimization of the task loss, a
condition that we call emph{consistency with the task loss}. In this work, we
propose another method for deriving differentiable surrogate losses that
provably meet this requirement. We focus on the broad class of models that
define a score for every input-output pair. Our idea is that this score can be
interpreted as an estimate of the task loss, and that the estimation error may
be used as a consistent surrogate loss. A distinct feature of such an approach
is that it defines the desirable value of the score for every input-output
pair. We use this property to design specialized surrogate losses for
Encoder-Decoder models often used for sequence prediction tasks. In our
experiment, we benchmark on the task of speech recognition. Using a new
surrogate loss instead of cross-entropy to train an Encoder-Decoder speech
recognizer brings a significant ~13% relative improvement in terms of Character
Error Rate (CER) in the case when no extra corpora are used for language
modeling.
|
DOC: Deep OCclusion Estimation From a Single Image | Recovering the occlusion relationships between objects is a fundamental human
visual ability which yields important information about the 3D world. In this
paper we propose a deep network architecture, called DOC, which acts on a
single image, detects object boundaries and estimates the border ownership
(i.e. which side of the boundary is foreground and which is background). We
represent occlusion relations by a binary edge map, to indicate the object
boundary, and an occlusion orientation variable which is tangential to the
boundary and whose direction specifies border ownership by a left-hand rule. We
train two related deep convolutional neural networks, called DOC, which exploit
local and non-local image cues to estimate this representation and hence
recover occlusion relations. In order to train and test DOC we construct a
large-scale instance occlusion boundary dataset using PASCAL VOC images, which
we call the PASCAL instance occlusion dataset (PIOD). This contains 10,000
images and hence is two orders of magnitude larger than existing occlusion
datasets for outdoor images. We test two variants of DOC on PIOD and on the
BSDS occlusion dataset and show they outperform state-of-the-art methods.
Finally, we perform numerous experiments investigating multiple settings of DOC
and transfer between BSDS and PIOD, which provides more insights for further
study of occlusion estimation.
|
Bayesian inference via rejection filtering | We provide a method for approximating Bayesian inference using rejection
sampling. We not only make the process efficient, but also dramatically reduce
the memory required relative to conventional methods by combining rejection
sampling with particle filtering. We also provide an approximate form of
rejection sampling that makes rejection filtering tractable in cases where
exact rejection sampling is not efficient. Finally, we present several
numerical examples of rejection filtering that show its ability to track time
dependent parameters in online settings and also benchmark its performance on
MNIST classification problems.
|
Unitary Evolution Recurrent Neural Networks | Recurrent neural networks (RNNs) are notoriously difficult to train. When the
eigenvalues of the hidden to hidden weight matrix deviate from absolute value
1, optimization becomes difficult due to the well studied issue of vanishing
and exploding gradients, especially when trying to learn long-term
dependencies. To circumvent this problem, we propose a new architecture that
learns a unitary weight matrix, with eigenvalues of absolute value exactly 1.
The challenge we address is that of parametrizing unitary matrices in a way
that does not require expensive computations (such as eigendecomposition) after
each weight update. We construct an expressive unitary weight matrix by
composing several structured matrices that act as building blocks with
parameters to be learned. Optimization with this parameterization becomes
feasible only when considering hidden states in the complex domain. We
demonstrate the potential of this architecture by achieving state of the art
results in several hard tasks involving very long-term dependencies.
|
On Binary Embedding using Circulant Matrices | Binary embeddings provide efficient and powerful ways to perform operations
on large scale data. However binary embedding typically requires long codes in
order to preserve the discriminative power of the input space. Thus binary
coding methods traditionally suffer from high computation and storage costs in
such a scenario. To address this problem, we propose Circulant Binary Embedding
(CBE) which generates binary codes by projecting the data with a circulant
matrix. The circulant structure allows us to use Fast Fourier Transform
algorithms to speed up the computation. For obtaining $k$-bit binary codes from
$d$-dimensional data, this improves the time complexity from $O(dk)$ to
$O(d\log{d})$, and the space complexity from $O(dk)$ to $O(d)$.
We study two settings, which differ in the way we choose the parameters of
the circulant matrix. In the first, the parameters are chosen randomly and in
the second, the parameters are learned using the data. For randomized CBE, we
give a theoretical analysis comparing it with binary embedding using an
unstructured random projection matrix. The challenge here is to show that the
dependencies in the entries of the circulant matrix do not lead to a loss in
performance. In the second setting, we design a novel time-frequency
alternating optimization to learn data-dependent circulant projections, which
alternatively minimizes the objective in original and Fourier domains. In both
the settings, we show by extensive experiments that the CBE approach gives much
better performance than the state-of-the-art approaches if we fix a running
time, and provides much faster computation with negligible performance
degradation if we fix the number of bits in the embedding.
|
Variance Reduction in SGD by Distributed Importance Sampling | Humans are able to accelerate their learning by selecting training materials
that are the most informative and at the appropriate level of difficulty. We
propose a framework for distributing deep learning in which one set of workers
search for the most informative examples in parallel while a single worker
updates the model on examples selected by importance sampling. This leads the
model to update using an unbiased estimate of the gradient which also has
minimum variance when the sampling proposal is proportional to the L2-norm of
the gradient. We show experimentally that this method reduces gradient variance
even in a context where the cost of synchronization across machines cannot be
ignored, and where the factors for importance sampling are not updated
instantly across the training set.
|
On the energy landscape of deep networks | We introduce "AnnealSGD", a regularized stochastic gradient descent algorithm
motivated by an analysis of the energy landscape of a particular class of deep
networks with sparse random weights. The loss function of such networks can be
approximated by the Hamiltonian of a spherical spin glass with Gaussian
coupling. While different from currently-popular architectures such as
convolutional ones, spin glasses are amenable to analysis, which provides
insights on the topology of the loss function and motivates algorithms to
minimize it. Specifically, we show that a regularization term akin to a
magnetic field can be modulated with a single scalar parameter to transition
the loss function from a complex, non-convex landscape with exponentially many
local minima, to a phase with a polynomial number of minima, all the way down
to a trivial landscape with a unique minimum. AnnealSGD starts training in the
relaxed polynomial regime and gradually tightens the regularization parameter
to steer the energy towards the original exponential regime. Even for
convolutional neural networks, which are quite unlike sparse random networks,
we empirically show that AnnealSGD improves the generalization error using
competitive baselines on MNIST and CIFAR-10.
|
Resiliency of Deep Neural Networks under Quantization | The complexity of deep neural network algorithms for hardware implementation
can be much lowered by optimizing the word-length of weights and signals.
Direct quantization of floating-point weights, however, does not show good
performance when the number of bits assigned is small. Retraining of quantized
networks has been developed to relieve this problem. In this work, the effects
of retraining are analyzed for a feedforward deep neural network (FFDNN) and a
convolutional neural network (CNN). The network complexity is controlled to
know their effects on the resiliency of quantized networks by retraining. The
complexity of the FFDNN is controlled by varying the unit size in each hidden
layer and the number of layers, while that of the CNN is done by modifying the
feature map configuration. We find that the performance gap between the
floating-point and the retrain-based ternary (+1, 0, -1) weight neural networks
exists with a fair amount in 'complexity limited' networks, but the discrepancy
almost vanishes in fully complex networks whose capability is limited by the
training data, rather than by the number of connections. This research shows
that highly complex DNNs have the capability of absorbing the effects of severe
weight quantization through retraining, but connection limited networks are
less resilient. This paper also presents the effective compression ratio to
guide the trade-off between the network size and the precision when the
hardware resource is limited.
|
The Variational Gaussian Process | Variational inference is a powerful tool for approximate inference, and it
has been recently applied for representation learning with deep generative
models. We develop the variational Gaussian process (VGP), a Bayesian
nonparametric variational family, which adapts its shape to match complex
posterior distributions. The VGP generates approximate posterior samples by
generating latent inputs and warping them through random non-linear mappings;
the distribution over random mappings is learned during inference, enabling the
transformed outputs to adapt to varying complexity. We prove a universal
approximation theorem for the VGP, demonstrating its representative power for
learning any model. For inference we present a variational objective inspired
by auto-encoders and perform black box inference over a wide class of models.
The VGP achieves new state-of-the-art results for unsupervised learning,
inferring models such as the deep latent Gaussian model and the recently
proposed DRAW.
|
Integrating Deep Features for Material Recognition | We propose a method for integration of features extracted using deep
representations of Convolutional Neural Networks (CNNs) each of which is
learned using a different image dataset of objects and materials for material
recognition. Given a set of representations of multiple pre-trained CNNs, we
first compute activations of features using the representations on the images
to select a set of samples which are best represented by the features. Then, we
measure the uncertainty of the features by computing the entropy of class
distributions for each sample set. Finally, we compute the contribution of each
feature to representation of classes for feature selection and integration. We
examine the proposed method on three benchmark datasets for material
recognition. Experimental results show that the proposed method achieves
state-of-the-art performance by integrating deep features. Additionally, we
introduce a new material dataset called EFMD by extending Flickr Material
Database (FMD). By the employment of the EFMD with transfer learning for
updating the learned CNN models, we achieve 84.0%+/-1.8% accuracy on the FMD
dataset which is close to human performance that is 84.9%.
|
Compression of Deep Convolutional Neural Networks for Fast and Low Power
Mobile Applications | Although the latest high-end smartphone has powerful CPU and GPU, running
deeper convolutional neural networks (CNNs) for complex tasks such as ImageNet
classification on mobile devices is challenging. To deploy deep CNNs on mobile
devices, we present a simple and effective scheme to compress the entire CNN,
which we call one-shot whole network compression. The proposed scheme consists
of three steps: (1) rank selection with variational Bayesian matrix
factorization, (2) Tucker decomposition on kernel tensor, and (3) fine-tuning
to recover accumulated loss of accuracy, and each step can be easily
implemented using publicly available tools. We demonstrate the effectiveness of
the proposed scheme by testing the performance of various compressed CNNs
(AlexNet, VGGS, GoogLeNet, and VGG-16) on the smartphone. Significant
reductions in model size, runtime, and energy consumption are obtained, at the
cost of small loss in accuracy. In addition, we address the important
implementation level issue on 1?1 convolution, which is a key operation of
inception module of GoogLeNet as well as CNNs compressed by our proposed
scheme.
|
Dueling Network Architectures for Deep Reinforcement Learning | In recent years there have been many successes of using deep representations
in reinforcement learning. Still, many of these applications use conventional
architectures, such as convolutional networks, LSTMs, or auto-encoders. In this
paper, we present a new neural network architecture for model-free
reinforcement learning. Our dueling network represents two separate estimators:
one for the state value function and one for the state-dependent action
advantage function. The main benefit of this factoring is to generalize
learning across actions without imposing any change to the underlying
reinforcement learning algorithm. Our results show that this architecture leads
to better policy evaluation in the presence of many similar-valued actions.
Moreover, the dueling architecture enables our RL agent to outperform the
state-of-the-art on the Atari 2600 domain.
|
Exponential Natural Particle Filter | Particle Filter algorithm (PF) suffers from some problems such as the loss of
particle diversity, the need for large number of particles, and the costly
selection of the importance density functions. In this paper, a novel
Exponential Natural Particle Filter (xNPF) is introduced to solve the above
problems. In this approach, a state transitional probability with the use of
natural gradient learning is proposed which balances exploration and
exploitation more robustly. The results show that xNPF converges much closer to
the true target states than the other state of the art particle filter.
|
Data Representation and Compression Using Linear-Programming
Approximations | We propose `Dracula', a new framework for unsupervised feature selection from
sequential data such as text. Dracula learns a dictionary of $n$-grams that
efficiently compresses a given corpus and recursively compresses its own
dictionary; in effect, Dracula is a `deep' extension of Compressive Feature
Learning. It requires solving a binary linear program that may be relaxed to a
linear program. Both problems exhibit considerable structure, their solution
paths are well behaved, and we identify parameters which control the depth and
diversity of the dictionary. We also discuss how to derive features from the
compressed documents and show that while certain unregularized linear models
are invariant to the structure of the compressed dictionary, this structure may
be used to regularize learning. Experiments are presented that demonstrate the
efficacy of Dracula's features.
|
Recurrent Gaussian Processes | We define Recurrent Gaussian Processes (RGP) models, a general family of
Bayesian nonparametric models with recurrent GP priors which are able to learn
dynamical patterns from sequential data. Similar to Recurrent Neural Networks
(RNNs), RGPs can have different formulations for their internal states,
distinct inference methods and be extended with deep structures. In such
context, we propose a novel deep RGP model whose autoregressive states are
latent, thereby performing representation and dynamical learning
simultaneously. To fully exploit the Bayesian nature of the RGP model we
develop the Recurrent Variational Bayes (REVARB) framework, which enables
efficient inference and strong regularization through coherent propagation of
uncertainty across the RGP layers and states. We also introduce a RGP extension
where variational parameters are greatly reduced by being reparametrized
through RNN-based sequential recognition models. We apply our model to the
tasks of nonlinear system identification and human motion modeling. The
promising obtained results indicate that our RGP model maintains its highly
flexibility while being able to avoid overfitting and being applicable even
when larger datasets are not available.
|
Recurrent Semi-supervised Classification and Constrained Adversarial
Generation with Motion Capture Data | We explore recurrent encoder multi-decoder neural network architectures for
semi-supervised sequence classification and reconstruction. We find that the
use of multiple reconstruction modules helps models generalize in a
classification task when only a small amount of labeled data is available,
which is often the case in practice. Such models provide useful high-level
representations of motions allowing clustering, searching and faster labeling
of new sequences. We also propose a new, realistic partitioning of a
well-known, high quality motion-capture dataset for better evaluations. We
further explore a novel formulation for future-predicting decoders based on
conditional recurrent generative adversarial networks, for which we propose
both soft and hard constraints for transition generation derived from desired
physical properties of synthesized future movements and desired animation
goals. We find that using such constraints allow to stabilize the training of
recurrent adversarial architectures for animation generation.
|
Modeling the Temporal Nature of Human Behavior for Demographics
Prediction | Mobile phone metadata is increasingly used for humanitarian purposes in
developing countries as traditional data is scarce. Basic demographic
information is however often absent from mobile phone datasets, limiting the
operational impact of the datasets. For these reasons, there has been a growing
interest in predicting demographic information from mobile phone metadata.
Previous work focused on creating increasingly advanced features to be modeled
with standard machine learning algorithms. We here instead model the raw mobile
phone metadata directly using deep learning, exploiting the temporal nature of
the patterns in the data. From high-level assumptions we design a data
representation and convolutional network architecture for modeling patterns
within a week. We then examine three strategies for aggregating patterns across
weeks and show that our method reaches state-of-the-art accuracy on both age
and gender prediction using only the temporal modality in mobile metadata. We
finally validate our method on low activity users and evaluate the modeling
assumptions.
|
L1 logistic regression as a feature selection step for training stable
classification trees for the prediction of severity criteria in imported
malaria | Multivariate classification methods using explanatory and predictive models
are necessary for characterizing subgroups of patients according to their risk
profiles. Popular methods include logistic regression and classification trees
with performances that vary according to the nature and the characteristics of
the dataset. In the context of imported malaria, we aimed at classifying
severity criteria based on a heterogeneous patient population. We investigated
these approaches by implementing two different strategies: L1 logistic
regression (L1LR) that models a single global solution and classification trees
that model multiple local solutions corresponding to discriminant subregions of
the feature space. For each strategy, we built a standard model, and a sparser
version of it. As an alternative to pruning, we explore a promising approach
that first constrains the tree model with an L1LR-based feature selection, an
approach we called L1LR-Tree. The objective is to decrease its vulnerability to
small data variations by removing variables corresponding to unstable local
phenomena. Our study is twofold: i) from a methodological perspective comparing
the performances and the stability of the three previous methods, i.e L1LR,
classification trees and L1LR-Tree, for the classification of severe forms of
imported malaria, and ii) from an applied perspective improving the actual
classification of severe forms of imported malaria by identifying more
personalized profiles predictive of several clinical criteria based on
variables dismissed for the clinical definition of the disease. The main
methodological results show that the combined method L1LR-Tree builds sparse
and stable models that significantly predicts the different severity criteria
and outperforms all the other methods in terms of accuracy.
|
Top-k Multiclass SVM | Class ambiguity is typical in image classification problems with a large
number of classes. When classes are difficult to discriminate, it makes sense
to allow k guesses and evaluate classifiers based on the top-k error instead of
the standard zero-one loss. We propose top-k multiclass SVM as a direct method
to optimize for top-k performance. Our generalization of the well-known
multiclass SVM is based on a tight convex upper bound of the top-k error. We
propose a fast optimization scheme based on an efficient projection onto the
top-k simplex, which is of its own interest. Experiments on five datasets show
consistent improvements in top-k accuracy compared to various baselines.
|
Top-N recommendations from expressive recommender systems | Normalized nonnegative models assign probability distributions to users and
random variables to items; see [Stark, 2015]. Rating an item is regarded as
sampling the random variable assigned to the item with respect to the
distribution assigned to the user who rates the item. Models of that kind are
highly expressive. For instance, using normalized nonnegative models we can
understand users' preferences as mixtures of interpretable user stereotypes,
and we can arrange properties of users and items in a hierarchical manner.
These features would not be useful if the predictive power of normalized
nonnegative models was poor. Thus, we analyze here the performance of
normalized nonnegative models for top-N recommendation and observe that their
performance matches the performance of methods like PureSVD which was
introduced in [Cremonesi et al., 2010]. We conclude that normalized nonnegative
models not only provide accurate recommendations but they also deliver (for
free) representations that are interpretable. We deepen the discussion of
normalized nonnegative models by providing further theoretical insights. In
particular, we introduce total variational distance as an operational
similarity measure, we discover scenarios where normalized nonnegative models
yield unique representations of users and items, we prove that the inference of
optimal normalized nonnegative models is NP-hard and finally, we discuss the
relationship between normalized nonnegative models and nonnegative matrix
factorization.
|
Scalable Gradient-Based Tuning of Continuous Regularization
Hyperparameters | Hyperparameter selection generally relies on running multiple full training
trials, with selection based on validation set performance. We propose a
gradient-based approach for locally adjusting hyperparameters during training
of the model. Hyperparameters are adjusted so as to make the model parameter
gradients, and hence updates, more advantageous for the validation cost. We
explore the approach for tuning regularization hyperparameters and find that in
experiments on MNIST, SVHN and CIFAR-10, the resulting regularization levels
are within the optimal regions. The additional computational cost depends on
how frequently the hyperparameters are trained, but the tested scheme adds only
30% computational overhead regardless of the model size. Since the method is
significantly less computationally demanding compared to similar gradient-based
approaches to hyperparameter optimization, and consistently finds good
hyperparameter values, it can be a useful tool for training neural network
models.
|
Hand Pose Estimation through Semi-Supervised and Weakly-Supervised
Learning | We propose a method for hand pose estimation based on a deep regressor
trained on two different kinds of input. Raw depth data is fused with an
intermediate representation in the form of a segmentation of the hand into
parts. This intermediate representation contains important topological
information and provides useful cues for reasoning about joint locations. The
mapping from raw depth to segmentation maps is learned in a
semi/weakly-supervised way from two different datasets: (i) a synthetic dataset
created through a rendering pipeline including densely labeled ground truth
(pixelwise segmentations); and (ii) a dataset with real images for which ground
truth joint positions are available, but not dense segmentations. Loss for
training on real images is generated from a patch-wise restoration process,
which aligns tentative segmentation maps with a large dictionary of synthetic
poses. The underlying premise is that the domain shift between synthetic and
real data is smaller in the intermediate representation, where labels carry
geometric and topological meaning, than in the raw input domain. Experiments on
the NYU dataset show that the proposed training method decreases error on
joints over direct regression of joints from depth data by 15.7%.
|
Sequence Level Training with Recurrent Neural Networks | Many natural language processing applications use language models to generate
text. These models are typically trained to predict the next word in a
sequence, given the previous words and some context such as an image. However,
at test time the model is expected to generate the entire sequence from
scratch. This discrepancy makes generation brittle, as errors may accumulate
along the way. We address this issue by proposing a novel sequence level
training algorithm that directly optimizes the metric used at test time, such
as BLEU or ROUGE. On three different tasks, our approach outperforms several
strong baselines for greedy generation. The method is also competitive when
these baselines employ beam search, while being several times faster.
|
Training CNNs with Low-Rank Filters for Efficient Image Classification | We propose a new method for creating computationally efficient convolutional
neural networks (CNNs) by using low-rank representations of convolutional
filters. Rather than approximating filters in previously-trained networks with
more efficient versions, we learn a set of small basis filters from scratch;
during training, the network learns to combine these basis filters into more
complex filters that are discriminative for image classification. To train such
networks, a novel weight initialization scheme is used. This allows effective
initialization of connection weights in convolutional layers composed of groups
of differently-shaped filters. We validate our approach by applying it to
several existing CNN architectures and training these networks from scratch
using the CIFAR, ILSVRC and MIT Places datasets. Our results show similar or
higher accuracy than conventional CNNs with much less compute. Applying our
method to an improved version of VGG-11 network using global max-pooling, we
achieve comparable validation accuracy using 41% less compute and only 24% of
the original VGG-11 model parameters; another variant of our method gives a 1
percentage point increase in accuracy over our improved VGG-11 model, giving a
top-5 center-crop validation accuracy of 89.7% while reducing computation by
16% relative to the original VGG-11 model. Applying our method to the GoogLeNet
architecture for ILSVRC, we achieved comparable accuracy with 26% less compute
and 41% fewer model parameters. Applying our method to a near state-of-the-art
network for CIFAR, we achieved comparable accuracy with 46% less compute and
55% fewer parameters.
|
Images Don't Lie: Transferring Deep Visual Semantic Features to
Large-Scale Multimodal Learning to Rank | Search is at the heart of modern e-commerce. As a result, the task of ranking
search results automatically (learning to rank) is a multibillion dollar
machine learning problem. Traditional models optimize over a few
hand-constructed features based on the item's text. In this paper, we introduce
a multimodal learning to rank model that combines these traditional features
with visual semantic features transferred from a deep convolutional neural
network. In a large scale experiment using data from the online marketplace
Etsy, we verify that moving to a multimodal representation significantly
improves ranking quality. We show how image features can capture fine-grained
style information not available in a text-only representation. In addition, we
show concrete examples of how image information can successfully disentangle
pairs of highly different items that are ranked similarly by a text-only model.
|
Data-Dependent Path Normalization in Neural Networks | We propose a unified framework for neural net normalization, regularization
and optimization, which includes Path-SGD and Batch-Normalization and
interpolates between them across two different dimensions. Through this
framework we investigate issue of invariance of the optimization, data
dependence and the connection with natural gradients.
|
Adding Gradient Noise Improves Learning for Very Deep Networks | Deep feedforward and recurrent networks have achieved impressive results in
many perception and language processing applications. This success is partially
attributed to architectural innovations such as convolutional and long
short-term memory networks. The main motivation for these architectural
innovations is that they capture better domain knowledge, and importantly are
easier to optimize than more basic architectures. Recently, more complex
architectures such as Neural Turing Machines and Memory Networks have been
proposed for tasks including question answering and general computation,
creating a new set of optimization challenges. In this paper, we discuss a
low-overhead and easy-to-implement technique of adding gradient noise which we
find to be surprisingly effective when training these very deep architectures.
The technique not only helps to avoid overfitting, but also can result in lower
training loss. This method alone allows a fully-connected 20-layer deep network
to be trained with standard gradient descent, even starting from a poor
initialization. We see consistent improvements for many complex models,
including a 72% relative reduction in error rate over a carefully-tuned
baseline on a challenging question-answering task, and a doubling of the number
of accurate binary multiplication models learned across 7,000 random restarts.
We encourage further application of this technique to additional complex modern
architectures.
|
Learning visual groups from co-occurrences in space and time | We propose a self-supervised framework that learns to group visual entities
based on their rate of co-occurrence in space and time. To model statistical
dependencies between the entities, we set up a simple binary classification
problem in which the goal is to predict if two visual primitives occur in the
same spatial or temporal context. We apply this framework to three domains:
learning patch affinities from spatial adjacency in images, learning frame
affinities from temporal adjacency in videos, and learning photo affinities
from geospatial proximity in image collections. We demonstrate that in each
case the learned affinities uncover meaningful semantic groupings. From patch
affinities we generate object proposals that are competitive with
state-of-the-art supervised methods. From frame affinities we generate movie
scene segmentations that correlate well with DVD chapter structure. Finally,
from geospatial affinities we learn groups that relate well to semantic place
categories.
|
GradNets: Dynamic Interpolation Between Neural Architectures | In machine learning, there is a fundamental trade-off between ease of
optimization and expressive power. Neural Networks, in particular, have
enormous expressive power and yet are notoriously challenging to train. The
nature of that optimization challenge changes over the course of learning.
Traditionally in deep learning, one makes a static trade-off between the needs
of early and late optimization. In this paper, we investigate a novel
framework, GradNets, for dynamically adapting architectures during training to
get the benefits of both. For example, we can gradually transition from linear
to non-linear networks, deterministic to stochastic computation, shallow to
deep architectures, or even simple downsampling to fully differentiable
attention mechanisms. Benefits include increased accuracy, easier convergence
with more complex architectures, solutions to test-time execution of batch
normalization, and the ability to train networks of up to 200 layers.
|
Online Sequence Training of Recurrent Neural Networks with Connectionist
Temporal Classification | Connectionist temporal classification (CTC) based supervised sequence
training of recurrent neural networks (RNNs) has shown great success in many
machine learning areas including end-to-end speech and handwritten character
recognition. For the CTC training, however, it is required to unroll (or
unfold) the RNN by the length of an input sequence. This unrolling requires a
lot of memory and hinders a small footprint implementation of online learning
or adaptation. Furthermore, the length of training sequences is usually not
uniform, which makes parallel training with multiple sequences inefficient on
shared memory models such as graphics processing units (GPUs). In this work, we
introduce an expectation-maximization (EM) based online CTC algorithm that
enables unidirectional RNNs to learn sequences that are longer than the amount
of unrolling. The RNNs can also be trained to process an infinitely long input
sequence without pre-segmentation or external reset. Moreover, the proposed
approach allows efficient parallel training on GPUs. For evaluation, phoneme
recognition and end-to-end speech recognition examples are presented on the
TIMIT and Wall Street Journal (WSJ) corpora, respectively. Our online model
achieves 20.7% phoneme error rate (PER) on the very long input sequence that is
generated by concatenating all 192 utterances in the TIMIT core test set. On
WSJ, a network can be trained with only 64 times of unrolling while sacrificing
4.5% relative word error rate (WER).
|
Unsupervised learning of object semantic parts from internal states of
CNNs by population encoding | We address the key question of how object part representations can be found
from the internal states of CNNs that are trained for high-level tasks, such as
object classification. This work provides a new unsupervised method to learn
semantic parts and gives new understanding of the internal representations of
CNNs. Our technique is based on the hypothesis that semantic parts are
represented by populations of neurons rather than by single filters. We propose
a clustering technique to extract part representations, which we call Visual
Concepts. We show that visual concepts are semantically coherent in that they
represent semantic parts, and visually coherent in that corresponding image
patches appear very similar. Also, visual concepts provide full spatial
coverage of the parts of an object, rather than a few sparse parts as is
typically found in keypoint annotations. Furthermore, We treat single visual
concept as part detector and evaluate it for keypoint detection using the
PASCAL3D+ dataset and for part detection using our newly annotated ImageNetPart
dataset. The experiments demonstrate that visual concepts can be used to detect
parts. We also show that some visual concepts respond to several semantic
parts, provided these parts are visually similar. Thus visual concepts have the
essential properties: semantic meaning and detection capability. Note that our
ImageNetPart dataset gives rich part annotations which cover the whole object,
making it useful for other part-related applications.
|
Data-dependent Initializations of Convolutional Neural Networks | Convolutional Neural Networks spread through computer vision like a wildfire,
impacting almost all visual tasks imaginable. Despite this, few researchers
dare to train their models from scratch. Most work builds on one of a handful
of ImageNet pre-trained models, and fine-tunes or adapts these for specific
tasks. This is in large part due to the difficulty of properly initializing
these networks from scratch. A small miscalibration of the initial weights
leads to vanishing or exploding gradients, as well as poor convergence
properties. In this work we present a fast and simple data-dependent
initialization procedure, that sets the weights of a network such that all
units in the network train at roughly the same rate, avoiding vanishing or
exploding gradients. Our initialization matches the current state-of-the-art
unsupervised or self-supervised pre-training methods on standard computer
vision tasks, such as image classification and object detection, while being
roughly three orders of magnitude faster. When combined with pre-training
methods, our initialization significantly outperforms prior work, narrowing the
gap between supervised and unsupervised pre-training.
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.