Recent work has shown how denoising and contractive autoencoders implicitly capture the structure of the data generating density, in the case where the corruption noise is Gaussian, the reconstruction error is the squared error, and the data is continuous-valued. This has led to various proposals for sampling from this implicitly learned density function, using Langevin and Metropolis-Hastings MCMC. However, it remained unclear how to connect the training procedure of regularized auto-encoders to the implicit estimation of the underlying data generating distribution when the data are discrete, or using other forms of corruption process and reconstruction errors. Another issue is the mathematical justification which is only valid in the limit of small corruption noise. We propose here a different attack on the problem, which deals with all these issues: arbitrary (but noisy enough) corruption, arbitrary reconstruction loss (seen as a log-likelihood), handling both discrete and continuous-valued variables, and removing the bias due to non-infinitesimal corruption noise (or non-infinitesimal contractive penalty).

In this paper, we study the challenging problem of tracking the trajectory of a moving object in a video with possibly very complex background. In contrast to most existing trackers which only learn the appearance of the tracked object online, we take a different approach, inspired by recent advances in deep learning architectures, by putting more emphasis on the (unsupervised) feature learning problem. Specifically, by using auxiliary natural images, we train a stacked denoising autoencoder offline to learn generic image features that are more robust against variations. This is then followed by knowledge transfer from offline training to the online tracking process. Online tracking involves a classification neural network which is constructed from the encoder part of the trained autoencoder as a feature extractor and an additional classification layer. Both the feature extractor and the classifier can be further tuned to adapt to appearance changes of the moving object. Comparison with the state-of-the-art trackers on some challenging benchmark video sequences shows that our deep learning tracker is very efficient as well as more accurate.

We introduce the multi-prediction deep Boltzmann machine (MP-DBM). The MP-DBM can be seen as a single probabilistic model trained to maximize a variational approximation to the generalized pseudolikelihood, or as a family of recurrent nets that share parameters and approximately solve different inference problems. Prior methods of training DBMs either do not perform well on classification tasks or require an initial learning pass that trains the DBM greedily, one layer at a time. The MP-DBM does not require greedy layerwise pretraining, and outperforms the standard DBM at classification, classification with missing inputs, and mean field prediction tasks.

Multilayer perceptrons (MLPs) or neural networks are popular models used for nonlinear regression and classification tasks. As regressors, MLPs model the conditional distribution of the predictor variables Y given the input variables X. However, this predictive distribution is assumed to be unimodal (e.g. Gaussian). For tasks such as structured prediction problems, the conditional distribution should be multimodal, forming one-to-many mappings. By using stochastic hidden variables rather than deterministic ones, Sigmoid Belief Nets (SBNs) can induce a rich multimodal distribution in the output space. However, previously proposed learning algorithms for SBNs are very slow and do not work well for real-valued data. In this paper, we propose a stochastic feedforward network with hidden layers having \emph{both deterministic and stochastic} variables. A new Generalized EM training procedure using importance sampling allows us to efficiently learn complicated conditional distributions. We demonstrate the superiority of our model to conditional Restricted Boltzmann Machines and Mixture Density Networks on synthetic datasets and on modeling facial expressions. Moreover, we show that latent features of our model improves classification and provide additional qualitative results on color images.

Time series often have a temporal hierarchy, with information that is spread out over multiple time scales. Common recurrent neural networks, however, do not explicitly accommodate such a hierarchy, and most research on them has been focusing on training algorithms rather than on their basic architecture. In this pa- per we study the effect of a hierarchy of recurrent neural networks on processing time series. Here, each layer is a recurrent network which receives the hidden state of the previous layer as input. This architecture allows us to perform hi- erarchical processing on difficult temporal tasks, and more naturally capture the structure of time series. We show that they reach state-of-the-art performance for recurrent networks in character-level language modelling when trained with sim- ple stochastic gradient descent. We also offer an analysis of the different emergent time scales.

As massively parallel computations have become broadly available with modern GPUs, deep architectures trained on very large datasets have risen in popularity. Discriminativelytrained convolutional neural networks, in particular, were recently shown to yield state-of-the-art performance in challenging image classification benchmarkssuch as ImageNet. However, elements of these architectures are similar to standard handcrafted representations used in computer vision. In this paper, we explore the extent of this analogy, proposing a version of the stateof-the-art Fishervector image encoding that can be stacked in multiple layers. This architecture significantly improves on standard Fisher vectors, and obtains competitive results with deep convolutional networks at a smaller computational learning cost. Our hybrid architecture allows us to assess how the performance of a conventional handcrafted image classification pipeline changes with increased depth. We also show that convolutional networks and Fisher vector encodings are complementary in the sense that their combination further improves the accuracy.

Hessian-free training has become a popular parallel second or- der optimization technique for Deep Neural Network training. This study aims at speeding up Hessian-free training, both by means of decreasing the amount of data used for training, as well as through reduction of the number of Krylov subspace solver iterations used for implicit estimation of the Hessian. In this paper, we develop an L-BFGS based preconditioning scheme that avoids the need to access the Hessian explicitly. Since L-BFGS cannot be regarded as a fixed-point iteration, we further propose the employment of flexible Krylov subspace solvers that retain the desired theoretical convergence guarantees of their conventional counterparts. Second, we propose a new sampling algorithm, which geometrically increases the amount of data utilized for gradient and Krylov subspace iteration calculations. On a 50-hr English Broadcast News task, we find that these methodologies provide roughly a 1.5x speed-up, whereas, on a 300-hr Switchboard task, these techniques provide over a 2.3x speedup, with no loss in WER. These results suggest that even further speed-up is expected, as problems scale and complexity grows.

Deep convolutional neural networks have recently achieved state-of-the-art performance on a number of image recognition benchmarks, including the ImageNet Large-Scale Visual Recognition Challenge (ILSVRC-2012). The winning model on the localization sub-task was a network that predicts a single bounding box and a confidence score for each object category in the image. Such a model captures the whole-image context around the objects but cannot handle multiple instances of the same object in the image without naively replicating the number of outputs for each instance. In this work, we propose a saliency-inspired neural network model for detection, which predicts a set of class-agnostic bounding boxes along with a single score for each box, corresponding to its likelihood of containing any object of interest. The model naturally handles a variable number of instances for each class and allows for cross-class generalization at the highest levels of the network. We are able to obtain competitive recognition performance on VOC2007 and ILSVRC2012, while using only the top few predicted locations in each image and a small number of neural network evaluations.

Recently, deep architectures, such as recurrent and recursive neural networks have been successfully applied to various natural language processing tasks. Inspired by bidirectional recurrent neural networks which use representations that summarize the past and future around an instance, we propose a novel architecture that aims to capture the structural information around an input, and use it to label instances. We apply our method to the task of opinion expression extraction, where we employ the binary parse tree of a sentence as the structure, and word vector representations as the initial representation of a single token. We conduct preliminary experiments to investigate its performance and compare it to the sequential approach.

Motivated by an important insight from neural science, we propose a new framework for understanding the success of the recently proposed "maxout" networks. The framework is based on encoding information on sparse pathways and recognizing the correct pathway at inference time. Elaborating further on this insight, we propose a novel deep network architecture, called "channel-out" network, which takes a much better advantage of sparse pathway encoding. In channel-out networks, pathways are not only formed a posteriori, but they are also actively selected according to the inference outputs from the lower layers. From a mathematical perspective, channel-out networks can represent a wider class of piece-wise continuous functions, thereby endowing the network with more expressive power than that of maxout networks. We test our channel-out networks on several well-known image classification benchmarks, setting new state-of-the-art performance on CIFAR-100 and STL-10, which represent some of the "harder" image classification benchmarks.