Machine learning algorithms are rarely parameter-free; whether via the properties of a regularizer, the hyperprior of a generative model, or the step size of a gradient-based optimization, learning procedures almost always require a set of high-level choices that significantly impact generalization performance. As a practitioner, one is usually able to specify the general framework of an inductive bias much more easily than the particular weighting that it should have relative to training data. As a result, these high-level parameters are often considered a nuisance, making it desirable to develop algorithms with as few of these "knobs" as possible. Another, more flexible take on this issue is to view the optimization of high-level parameters as a procedure to be automated. Specifically, we could view such tuning as the optimization of an unknown black-box function that reflects generalization performance and invoke algorithms developed for such problems. These optimization problems have a somewhat different flavor than the low-level objectives one often encounters as part of a training procedure: here function evaluations are very expensive, as they involve running the primary machine learning algorithm to completion. In this setting where function evaluations are expensive, it is desirable to spend computational time making better choices about where to seek the best parameters. Bayesian optimization (Mockus et al., 1978) provides an elegant approach and has been shown to outperform other state of the art global optimization algorithms on a number of challenging optimization benchmark functions (Jones, 2001).

Extractive style query oriented multi document summariza tion generates the summary by extracting a proper set of sentences from multiple documents based on the pre given query. This paper proposes a novel multi document summa rization framework via deep learning model. This uniform framework consists of three parts: concepts extraction, summary generation, and reconstruction validation, which work together to achieve the largest coverage of the docu ments content. A new query oriented extraction technique is proposed to concentrate distributed information to hidden units layer by layer. Then, the whole deep architecture is fi ne tuned by minimizing the information loss of reconstruc tion validation. According to the concentrated information, dynamic programming is used to seek most informative set of sentences as the summary. Experiments on three bench mark datasets demonstrate the effectiveness of the proposed framework and algorithms.

The restricted Boltzmann machine (RBM) is a flexible tool for modeling complex data, however there have been significant computational difficulties in using RBMs to model high-dimensional multinomial observations. In natural language processing applications, words are naturally modeled by K-ary discrete distributions, where K is determined by the vocabulary size and can easily be in the hundreds of thousands. The conventional approach to training RBMs on word observations is limited because it requires sampling the states of K-way softmax visible units during block Gibbs updates, an operation that takes time linear in K. In this work, we address this issue by employing a more general class of Markov chain Monte Carlo operators on the visible units, yielding updates with computational complexity independent of K. We demonstrate the success of our approach by training RBMs on hundreds of millions of word n-grams using larger vocabularies than previously feasible and using the learned features to improve performance on chunking and sentiment classification tasks, achieving state-of-the-art results on the latter.

Echo state network (ESN) is viewed as a temporal non-orthogonal expansion with pseudo-random parameters. Such expansions naturally give rise to regressors of various relevance to a teacher output. We illustrate that often only a certain amount of the generated echo-regressors effectively explain the variance of the teacher output and also that sole local regularization is not able to provide in-depth information concerning the importance of the generated regressors. The importance is therefore determined by a joint calculation of the individual variance contributions and Bayesian relevance using locally regularized orthogonal forward regression (LROFR) algorithm. This information can be advantageously used in a variety of ways for an in-depth analysis of an ESN structure and its state-space parameters in relation to the unknown dynamics of the underlying problem. We present locally regularized linear readout built using LROFR. The readout may have a different dimensionality than an ESN model itself, and besides improving robustness and accuracy of an ESN it relates the echo-regressors to different features of the training data and may determine what type of an additional readout is suitable for a task at hand. Moreover, as flexibility of the linear readout has limitations and might sometimes be insufficient for certain tasks, we also present a radial basis function (RBF) readout built using LROFR. It is a flexible and parsimonious readout with excellent generalization abilities and is a viable alternative to readouts based on a feed-forward neural network (FFNN) or an RBF net built using relevance vector machine (R VM). Introduction ESNs are a novel class of recurrent neural networks (RNN) [1]. Their easy construction and simple training procedure are appealing and have attracted the attention of many researchers. Vector function f is applied element-wise to its arguments. The most common choice forf is either a vector of sigmoid or identity functions. The expansion is carried out so that diverse echoes of an input and teacher signal are generated (hence the name echo-state). This diversity, which should appropriately "explain" a variance of a teacher signal, is the key to the successful training of an ESN.

Recent work suggests that some auto-encoder variants do a good job of capturing the local manifold structure of the unknown data generating density. This paper contributes to the mathematical understanding of this phenomenon and helps define better justified sampling algorithms for deep learning based on auto-encoder variants. We consider an MCMC where each step samples from a Gaussian whose mean and covariance matrix depend on the previous state, defines through its asymptotic distribution a target density. First, we show that good choices (in the sense of consistency) for these mean and covariance functions are the local expected value and local covariance under that target density. Then we show that an auto-encoder with a contractive penalty captures estimators of these local moments in its reconstruction function and its Jacobian. A contribution of this work is thus a novel alternative to maximum-likelihood density estimation, which we call local moment matching. It also justifies a recently proposed sampling algorithm for the Contractive Auto-Encoder and extends it to the Denoising Auto-Encoder.

Visual perception is a challenging problem in part due to illumination variations. A possible solution is to first estimate an illumination invariant representation before using it for recognition. The object albedo and surface normals are examples of such representations. In this paper, we introduce a multilayer generative model where the latent variables include the albedo, surface normals, and the light source. Combining Deep Belief Nets with the Lambertian reflectance assumption, our model can learn good priors over the albedo from 2D images. Illumination variations can be explained by changing only the lighting latent variable in our model. By transferring learned knowledge from similar objects, albedo and surface normals estimation from a single image is possible in our model. Experiments demonstrate that our model is able to generalize as well as improve over standard baselines in one-shot face recognition.

The contractive auto-encoder learns a representation of the input data that captures the local manifold structure around each data point, through the leading singular vectors of the Jacobian of the transformation from input to representation. The corresponding singular values specify how much local variation is plausible in directions associated with the corresponding singular vectors, while remaining in a high-density region of the input space. This paper proposes a procedure for generating samples that are consistent with the local structure captured by a contractive auto-encoder. The associated stochastic process defines a distribution from which one can sample, and which experimentally appears to converge quickly and mix well between modes, compared to Restricted Boltzmann Machines and Deep Belief Networks. The intuitions behind this procedure can also be used to train the second layer of contraction that pools lower-level features and learns to be invariant to the local directions of variation discovered in the first layer. We show that this can help learn and represent invariances present in the data and improve classification error.

Learning invariant representations is an important problem in machine learning and pattern recognition. In this paper, we present a novel framework of transformation-invariant feature learning by incorporating linear transformations into the feature learning algorithms. For example, we present the transformation-invariant restricted Boltzmann machine that compactly represents data by its weights and their transformations, which achieves invariance of the feature representation via probabilistic max pooling. In addition, we show that our transformation-invariant feature learning framework can also be extended to other unsupervised learning methods, such as autoencoders or sparse coding. We evaluate our method on several image classification benchmark datasets, such as MNIST variations, CIFAR-10, and STL-10, and show competitive or superior classification performance when compared to the state-of-the-art. Furthermore, our method achieves state-of-the-art performance on phone classification tasks with the TIMIT dataset, which demonstrates wide applicability of our proposed algorithms to other domains.

We consider the problem of object recognition with a large number of classes. In order to overcome the low amount of labeled examples available in this setting, we introduce a new feature learning and extraction procedure based on a factor model we call spike-and-slab sparse coding (S3C). Prior work on S3C has not prioritized the ability to exploit parallel architectures and scale S3C to the enormous problem sizes needed for object recognition. We present a novel inference procedure for appropriate for use with GPUs which allows us to dramatically increase both the training set size and the amount of latent factors that S3C may be trained with. We demonstrate that this approach improves upon the supervised learning capabilities of both sparse coding and the spike-and-slab Restricted Boltzmann Machine (ssRBM) on the CIFAR-10 dataset. We use the CIFAR-100 dataset to demonstrate that our method scales to large numbers of classes better than previous methods. Finally, we use our method to win the NIPS 2011 Workshop on Challenges In Learning Hierarchical Models? Transfer Learning Challenge.

We investigate the problem of modeling symbolic sequences of polyphonic music in a completely general piano-roll representation. We introduce a probabilistic model based on distribution estimators conditioned on a recurrent neural network that is able to discover temporal dependencies in high-dimensional sequences. Our approach outperforms many traditional models of polyphonic music on a variety of realistic datasets. We show how our musical language model can serve as a symbolic prior to improve the accuracy of polyphonic transcription.