We propose a deep convolutional neural network architecture codenamed Inception that achieves the new state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC14). The main hallmark of this architecture is the improved utilization of the computing resources inside the network. By a carefully crafted design, we increased the depth and width of the network while keeping the computational budget constant. To optimize quality, the architectural decisions were based on the Hebbian principle and the intuition of multi-scale processing. One particular incarnation used in our submission for ILSVRC14 is called GoogLeNet, a 22 layers deep network, the quality of which is assessed in the context of classification and detection.

/pdf/going-deeper-with-convolutions-1yobw2o2ds.pdf

Going deeper with convolutions

Deep learning is a form of machine learning that enables computers to learn from experience and understand the world in terms of a hierarchy of concepts. Because the computer gathers knowledge from experience, there is no need for a human computer operator to formally specify all the knowledge that the computer needs. The hierarchy of concepts allows the computer to learn complicated concepts by building them out of simpler ones; a graph of these hierarchies would be many layers deep. This book introduces a broad range of topics in deep learning. The text offers mathematical and conceptual background, covering relevant concepts in linear algebra, probability theory and information theory, numerical computation, and machine learning. It describes deep learning techniques used by practitioners in industry, including deep feedforward networks, regularization, optimization algorithms, convolutional networks, sequence modeling, and practical methodology; and it surveys such applications as natural language processing, speech recognition, computer vision, online recommendation systems, bioinformatics, and videogames. Finally, the book offers research perspectives, covering such theoretical topics as linear factor models, autoencoders, representation learning, structured probabilistic models, Monte Carlo methods, the partition function, approximate inference, and deep generative models. Deep Learning can be used by undergraduate or graduate students planning careers in either industry or research, and by software engineers who want to begin using deep learning in their products or platforms. A website offers supplementary material for both readers and instructors.

Deep Learning

We describe latent Dirichlet allocation (LDA), a generative probabilistic model for collections of discrete data such as text corpora. LDA is a three-level hierarchical Bayesian model, in which each item of a collection is modeled as a finite mixture over an underlying set of topics. Each topic is, in turn, modeled as an infinite mixture over an underlying set of topic probabilities. In the context of text modeling, the topic probabilities provide an explicit representation of a document. We present efficient approximate inference techniques based on variational methods and an EM algorithm for empirical Bayes parameter estimation. We report results in document modeling, text classification, and collaborative filtering, comparing to a mixture of unigrams model and the probabilistic LSI model.

/pdf/latent-dirichlet-allocation-4gmscb06en.pdf

Latent dirichlet allocation

We present a new technique called “t-SNE” that visualizes high-dimensional data by giving each datapoint a location in a two or three-dimensional map. The technique is a variation of Stochastic Neighbor Embedding (Hinton and Roweis, 2002) that is much easier to optimize, and produces significantly better visualizations by reducing the tendency to crowd points together in the center of the map. t-SNE is better than existing techniques at creating a single map that reveals structure at many different scales. This is particularly important for high-dimensional data that lie on several different, but related, low-dimensional manifolds, such as images of objects from multiple classes seen from multiple viewpoints. For visualizing the structure of very large datasets, we show how t-SNE can use random walks on neighborhood graphs to allow the implicit structure of all of the data to influence the way in which a subset of the data is displayed. We illustrate the performance of t-SNE on a wide variety of datasets and compare it with many other non-parametric visualization techniques, including Sammon mapping, Isomap, and Locally Linear Embedding. The visualizations produced by t-SNE are significantly better than those produced by the other techniques on almost all of the datasets.

/pdf/visualizing-data-using-t-sne-2w1n304iq0.pdf

Visualizing Data using t-SNE

We propose a generative model for text and other collections of discrete data that generalizes or improves on several previous models including naive Bayes/unigram, mixture of unigrams [6], and Hof-mann's aspect model, also known as probabilistic latent semantic indexing (pLSI) [3]. In the context of text modeling, our model posits that each document is generated as a mixture of topics, where the continuous-valued mixture proportions are distributed as a latent Dirichlet random variable. Inference and learning are carried out efficiently via variational algorithms. We present empirical results on applications of this model to problems in text modeling, collaborative filtering, and text classification.

/pdf/latent-dirichlet-allocation-rkwbqihm3p.pdf

Latent Dirichlet Allocation

We propose a statistical mechanical framework for the modeling of discrete time series. Maximum likelihood estimation is done via Boltzmann learning in one-dimensional networks with tied weights. We call these networks Boltzmann chains and show that they contain hidden Markov models (HMMs) as a special case. Our framework also motivates new architectures that address particular shortcomings of HMMs. We look at two such architectures: parallel chains that model feature sets with disparate time scales, and looped networks that model long-term dependencies between hidden states. For these networks, we show how to implement the Boltzmann learning rule exactly, in polynomial time, without resort to simulated or mean-field annealing. The necessary computations are done by exact decimation procedures from statistical mechanics.

/pdf/boltzmann-chains-and-hidden-markov-models-4bwctkiy1x.pdf

Boltzmann Chains and Hidden Markov Models

Many different high dimensional data sets are characterized by the same underlying modes of variability. When these modes of variability are continuous and few in number, they can be viewed as parameterizing a low dimensional manifold. The manifold provides a compact shared representation of the data, suggesting correspondences between the high dimensional examples from different data sets. These correspondences, though naturally induced by the underlying manifold, are difficult to learn using traditional methods in supervised learning. In this paper, we generalize three methods in unsupervised learning—principal components analysis, factor analysis, and locally linear embedding— to discover subspaces and manifolds that provide common low dimensional representations of different high dimensional data sets. We use the shared representations discovered by these algorithms to put high dimensional examples from different data sets into correspondence. Finally, we show that a notion of “self-correspondence” between examples in the same data set can be used to improve the performance of these algorithms on small data sets. The algorithms are demonstrated on images and text.

https://repository.upenn.edu/cgi/viewcontent.cgi?article=1131&context=ese_papers

Learning High Dimensional Correspondences from Low Dimensional Manifolds

Hidden Markov models (HMMs) for automatic speech recognition rely on high dimensional feature vectors to summarize the short-time properties of speech. Correlations between features can arise when the speech signal is nonstationary or corrupted by noise. We investigate how to model these correlations using factor analysis, a statistical method for dimensionality reduction. Factor analysis uses a small number of parameters to model the covariance structure of high dimensional data. These parameters can be chosen in two ways: (1) to maximize the likelihood of observed speech signals, or (2) to minimize the number of classification errors. We derive an expectation-maximization (EM) algorithm for maximum likelihood estimation and a gradient descent algorithm for improved class discrimination. Speech recognizers are evaluated on two tasks, one small-sized vocabulary (connected alpha-digits) and one medium-sized vocabulary (New Jersey town names). We find that modeling feature correlations by factor analysis leads to significantly increased likelihoods and word accuracies. Moreover, the rate of improvement with model size often exceeds that observed in conventional HMM's.

/pdf/maximum-likelihood-and-minimum-classification-error-factor-9tehhicipk.pdf

Maximum likelihood and minimum classification error factor analysis for automatic speech recognition

The paper explores a generalization of conditional random elds (CRFs) in which binary stochastic hidden units appear between the data and the labels. Hidden-unit CRFs are potentially more powerful than standard CRFs because they can represent nonlinear dependencies at each frame. The hidden units in these models also learn to discover latent distributed structure in the data that improves classication. We derive ecient algorithms for inference and learning in these models by observing that the hidden units are conditionally independent given the data and the labels. Finally, we show that hiddenunit CRFs perform well in experiments on a range of tasks, including optical character recognition, text classication, protein structure prediction, and part-of-speech tagging.

/pdf/hidden-unit-conditional-random-fields-23zvgncawx.pdf

Hidden-Unit Conditional Random Fields

In this paper we compare three frameworks for discriminative training of continuous-density hidden Markov models (CD-HMMs). Specifically, we compare two popular frameworks, based on conditional maximum likelihood (CML) and minimum classification error (MCE), to a new framework based on margin maximization. Unlike CML and MCE, our formulation of large margin training explicitly penalizes incorrect decodings by an amount proportional to the number of mislabeled hidden states. It also leads to a convex optimization over the parameter space of CD-HMMs, thus avoiding the problem of spurious local minima. We used discriminatively trained CD-HMMs from all three frameworks to build phonetic recognizers on the TIMIT speech corpus. The different recognizers employed exactly the same acoustic front end and hidden state space, thus enabling us to isolate the effect of different cost functions, parameterizations, and numerical optimizations. Experimentally, we find that our framework for large margin training yields significantly lower error rates than both CML and MCE training.

/pdf/comparison-of-large-margin-training-to-other-discriminative-2wvz8rqmoy.pdf

Lawrence K. Saul

Papers

Boltzmann Chains and Hidden Markov Models

Learning High Dimensional Correspondences from Low Dimensional Manifolds

Maximum likelihood and minimum classification error factor analysis for automatic speech recognition

Hidden-Unit Conditional Random Fields

Comparison of Large Margin Training to Other Discriminative Methods for Phonetic Recognition by Hidden Markov Models