In this paper, we consider a finite-horizon Markov decision process (MDP) for which the objective at each stage is to minimize a quantile-based risk measure (QBRM) of the sequence of future costs; we call the overall objective a dynamic quantile-based risk measure (DQBRM). In particular, we consider optimizing dynamic risk measures where the one-step risk measures are QBRMs, a class of risk measures that includes the popular value at risk (VaR) and the conditional value at risk (CVaR). Although there is considerable theoretical development of risk-averse MDPs in the literature, the computational challenges have not been explored as thoroughly. We propose data-driven and simulation-based approximate dynamic programming (ADP) algorithms to solve the risk-averse sequential decision problem. We address the issue of inefficient sampling for risk applications in simulated settings and present a procedure, based on importance sampling, to direct samples toward the "risky region" as the ADP algorithm progresses. Finally, we show numerical results of our algorithms in the context of an application involving risk-averse bidding for energy storage.

End-to-end automatic speech recognition system implemented in TensorFlow. If you want to replace feed dict operation with Tensorflow multi-thread and fifoqueue input pipeline, you can refer to my repo TensorFlow-Input-Pipeline for more example codes. My own practices prove that fifoqueue input pipeline would improve the training speed in some time. If you want to look the history of speech recognition, I have collected the significant papers since 1981 in the ASR field. I will update it every week to add new papers, including speech recognition, speech synthesis and language modelling.

In many applications, data come with a natural ordering. This ordering can often induce local dependence among nearby variables. However, in complex data, the width of this dependence may vary, making simple assumptions such as a constant neighborhood size unrealistic. We propose a framework for learning this local dependence based on estimating the inverse of the Cholesky factor of the covariance matrix. Penalized maximum likelihood estimation of this matrix yields a simple regression interpretation for local dependence in which variables are predicted by their neighbors. Our proposed method involves solving a convex, penalized Gaussian likelihood problem with a hierarchical group lasso penalty. The problem decomposes into independent subproblems which can be solved efficiently in parallel using first-order methods. Our method yields a sparse, symmetric, positive definite estimator of the precision matrix, encoding a Gaussian graphical model. We derive theoretical results not found in existing methods attaining this structure. In particular, our conditions for signed support recovery and estimation consistency rates in multiple norms are as mild as those in a regression problem. Empirical results show our method performing favorably compared to existing methods. We apply our method to genomic data to flexibly model linkage disequilibrium. Our method is also applied to improve the performance of discriminant analysis in sound recording classification.

Artificial intelligence (AI), machine learning (ML), and robots are the sights and sounds of science fiction books and movies. Isaac Asimov's Three Laws of Robotics, first introduced in the 1942 short story "Runaround," became the backbone for his novel I, Robot and its film adaptation (Figure 1). Although we are still far away from achieving what movie producers and sci-fi writers have envisioned, the state of AI and ML has progressed significantly. AI software has also been in use for decades but advances in ML, including the use of deep neural networks (DNNs), are making headlines in application areas like self-driving cars. The movie I, Robot has robots that should be following Asimov's Three Laws of Robotics.

A few weeks ago at the DALI Theory of GANs workshop we had a great discussion about what GANs are even useful for. Pretty much everybody agreed that generating random images from a model is not really our goal. We either want to use GANs to train conditional probabilistic models (like we do for image super-resolution or speech synthesis, or something along those lines), or as a means of unsupervised representation learning. Indeed, many papers examine the latent space representations that GANs learn. But the elephant in the room is that nobody really agrees on what unsupervised representation learning really means, and why any GAN variant should be any better or worse at it than others, whether GANs or VAEs are better for that.

Online reviews provided by consumers are a valuable asset for e-Commerce platforms, influencing potential consumers in making purchasing decisions. However, these reviews are of varying quality, with the useful ones buried deep within a heap of non-informative reviews. In this work, we attempt to automatically identify review quality in terms of its helpfulness to the end consumers. In contrast to previous works in this domain exploiting a variety of syntactic and community-level features, we delve deep into the semantics of reviews as to what makes them useful, providing interpretable explanation for the same. We identify a set of consistency and semantic factors, all from the text, ratings, and timestamps of user-generated reviews, making our approach generalizable across all communities and domains. We explore review semantics in terms of several latent factors like the expertise of its author, his judgment about the fine-grained facets of the underlying product, and his writing style. These are cast into a Hidden Markov Model -- Latent Dirichlet Allocation (HMM-LDA) based model to jointly infer: (i) reviewer expertise, (ii) item facets, and (iii) review helpfulness. Large-scale experiments on five real-world datasets from Amazon show significant improvement over state-of-the-art baselines in predicting and ranking useful reviews.

Current recommender systems exploit user and item similarities by collaborative filtering. Some advanced methods also consider the temporal evolution of item ratings as a global background process. However, all prior methods disregard the individual evolution of a user's experience level and how this is expressed in the user's writing in a review community. In this paper, we model the joint evolution of user experience, interest in specific item facets, writing style, and rating behavior. This way we can generate individual recommendations that take into account the user's maturity level (e.g., recommending art movies rather than blockbusters for a cinematography expert). As only item ratings and review texts are observables, we capture the user's experience and interests in a latent model learned from her reviews, vocabulary and writing style. We develop a generative HMM-LDA model to trace user evolution, where the Hidden Markov Model (HMM) traces her latent experience progressing over time -- with solely user reviews and ratings as observables over time. The facets of a user's interest are drawn from a Latent Dirichlet Allocation (LDA) model derived from her reviews, as a function of her (again latent) experience level. In experiments with five real-world datasets, we show that our model improves the rating prediction over state-of-the-art baselines, by a substantial margin. We also show, in a use-case study, that our model performs well in the assessment of user experience levels.

We propose a new reinforcement learning algorithm for partially observable Markov decision processes (POMDP) based on spectral decomposition methods. While spectral methods have been previously employed for consistent learning of (passive) latent variable models such as hidden Markov models, POMDPs are more challenging since the learner interacts with the environment and possibly changes the future observations in the process. We devise a learning algorithm running through epochs, in each epoch we employ spectral techniques to learn the POMDP parameters from a trajectory generated by a fixed policy. At the end of the epoch, an optimization oracle returns the optimal memoryless planning policy which maximizes the expected reward based on the estimated POMDP model. We prove an order-optimal regret bound with respect to the optimal memoryless policy and efficient scaling with respect to the dimensionality of observation and action spaces.

As manufacturing processes become increasingly automated, so should tool condition monitoring (TCM) as it is impractical to have human workers monitor the state of the tools continuously. Tool condition is crucial to ensure the good quality of products: Worn tools affect not only the surface quality but also the dimensional accuracy, which means higher reject rate of the products. Therefore, there is an urgent need to identify tool failures before it occurs on the fly. While various versions of intelligent tool condition monitoring have been proposed, most of them suffer from a cognitive nature of traditional machine learning algorithms. They focus on the how to learn process without paying attention to other two crucial issues: what to learn, and when to learn. The what to learn and the when to learn provide self regulating mechanisms to select the training samples and to determine time instants to train a model. A novel tool condition monitoring approach based on a psychologically plausible concept, namely the metacognitive scaffolding theory, is proposed and built upon a recently published algorithm, recurrent classifier (rClass). The learning process consists of three phases: what to learn, how to learn, when to learn and makes use of a generalized recurrent network structure as a cognitive component. Experimental studies with real-world manufacturing data streams were conducted where rClass demonstrated the highest accuracy while retaining the lowest complexity over its counterparts.

This study introduces statistical boosting for capture-mark-recapture (CMR) models. It is a shrinkage estimator that constrains the complexity of a CMR model in order to promote automatic variable-selection and avoid over-fitting. I discuss the philosophical similarities between boosting and AIC model-selection, and show through simulations that a boosted Cormack-Jolly-Seber model often out-performs AICc methods, in terms of estimating survival and abundance, yet yields qualitatively similar estimates. This new boosted CMR framework is highly extensible and could provide a rich, unified framework for addressing many topics in CMR, such as non-linear effects (splines and CART-like trees), individual-heterogeneity, and spatial components.