This paper introduces a method for solving a general class of structured prediction problems. The method trains a neural network to construct an output as a deterministic function of the real input and a sample from some noise source. Entropy in the noise source becomes entropy in the output distribution. Mismatch between the model distribution and true predictive distribution is measured using a strictly proper scoring rule, a la Gneiting and Raftery (JASA 2007). One thing that concerns me about the proposed approach is whether the "expected score" that's used for measuring dissimilarity between the model predictions and the true predictive distribution provides a strong learning signal. Especially in the minibatch setting, I'd be worried about variance in the gradient wiping out information about subtle mismatch between the model and true distributions.

For the valuable problem of large-scale and sparse stochastic inference on simplex, the authors proposed a novel Stochastic gradient Markov chain Monte Carlo (SGMCMC) method, which is based on the Cox-Ingersoll-Ross (CIR) process. Compared with the commonly-used Langevin diffusion within the SGMCMC community, the CIR process (i) is closely related to the flexible Gamma distribution, and therefore more suitable for inferring a Dirichlet distribution on simplex, since a Dirichlet distribution is just the normalization of Gamma distributions; (ii) CIR has no discretization error, which is shown to be a clear advantage over the Langevin diffusion on simplex inference. Besides, the author proved that the proposed SCIR method is asymptotically unbiased, and has improved performance over other SGMCMC method on sparse simplex problem via two experiments, namely inferring a LDA on a dataset of scraped Wikipedia documents and inferring a Bayesian nonparametric mixture model on Microsoft user dataset. I think the quality is good; the presentation is clear; as far as I know the proposed technique is original and of great significance. Therefore I vote for acceptance. However, the experiments are okay, but not strong.

I like working with textual data. As for Computer Vision, in NLP nowadays there are a lot of ready accessible resources and opensource projects, which we can directly download or consume. Some of them are realy cool and permit us to speed up and bring to another level our projects. The most important thing we must not forgotten is that all these instruments aren't magic. Some of them declare high performances but they are nothing if we don't allow them to make the best.

Ask Me Anything: Dynamic Memory Networks for Natural Language Processing, Ankit Kumar, Ozan Irsoy, Peter Ondruska, Mohit Iyyer, James Bradbury, Ishaan Gulrajani, Victor Zhong, Romain Paulus, Richard Socher The 33rd International Conference on Machine Learning (ICML 2016).

Rich Sutton is old school king of RL.

The Conference and Workshop on Neural Information Processing Systems (NIPS) is a machine learning and computational neuroscience conference held every December. The conference is a single track meeting that includes invited talks as well as oral and poster presentations of refereed papers, followed by parallel-track workshops that up to 2013 were held at ski resorts. According to Microsoft Academic Search, NIPS is the top conference on machine learning. US tops the chart and has been the top research contributor with the highest number of accepts. UK stands second with an marginal increase in paper accepts from 2014, whereas India and China both dropped, with lesser number of accepts in 2015.

Integration is the central numerical operation required for Bayesian machine learning (in the form of marginalization and conditioning). Sampling algorithms still abound in this area, although it has long been known that Monte Carlo methods are fundamentally sub-optimal. The challenges for the development of better performing integration methods are mostly algorithmic. Moreover, recent algorithms have begun to outperform MCMC and its siblings, in wall-clock time, on realistic problems from machine learning. A community website for probabilistic numerics can be found at http://probabilistic-numerics.org.

This volume is a collection of contributions from the 5th Workshop on Machine Learning and Interpretation in Neuroimaging (MLINI) at the Neural Information Processing Systems (NIPS 2015) conference. Modern multivariate statistical methods developed in the rapidly growing field of machine learning are being increasingly applied to various problems in neuroimaging, from cognitive state detection to clinical diagnosis and prognosis. Multivariate pattern analysis methods are designed to examine complex relationships between high-dimensional signals, such as brain images, and outcomes of interest, such as the category of a stimulus, a type of a mental state of a subject, or a specific mental disorder. Such techniques are in contrast with the traditional mass-univariate approaches that dominated neuroimaging in the past and treated each individual imaging measurement in isolation. We believe that machine learning has a prominent role in shaping how questions in neuroscience are framed, and that the machine-learning mind set is now entering modern psychology and behavioral studies. It is also equally important that practical applications in these fields motivate a rapidly evolving line or research in the machine learning community. In parallel, there is an intense interest in learning more about brain function in the context of rich naturalistic environments and scenes. Efforts to go beyond highly specific paradigms that pinpoint a single function, towards schemes for measuring the interaction with natural and more varied scene are made. The goal of the workshop is to pinpoint the most pressing issues and common challenges across the neuroscience, neuroimaging, psychology and machine learning fields, and to sketch future directions and open questions in the light of novel methodology.

If you think of each day as a conference crammed into a day, you get a good flavor of things. Here are some of the interesting things I saw.

NIPS 2015 was bigger than ever, literally: at circa 3700 attendees this was roughly twice as many attendees as last year, which in turn was roughly twice as many as the previous year. This is clearly unsustainable, but given the frenzied level of vendor and recruiting activities, perhaps there is room to grow. The main conference is single track, however, and already 3 days long: so even more action is moving to the poster sessions, which along with the workshops creates the feel of a diverse collection of smaller conferences. Obviously, my view of the action will be highly incomplete and biased towards my own interests. Reinforcement learning continues to ascend, extending the enthusiasm and energy from ICML.