Technical quality: This paper tackles the problem of inference and learning of factored HMMs on large sequences and with large latent dimensionality. The primary contribution of the paper is integrating several existing approaches together to enable large-scale learning of FHMMs without a loss in modeling performance. The technical details of the components of the approach (the bivariate Gaussian copula variation posterior, the recognition network, the SVI learning approach) appear to be technically correct. The experimentation touches on the correct points including the accuracy of the learned models and the scalability of the proposed approach. The accuracy of the learned models is assessed using log likelihood on held-out test data. The experiments show that the model performs similarly to the SMF approach on both simulated and real (the Bach Corals) data.

This is a very good and technically sound paper, containing a significant amount of material. The theoretical investigation of the properties of alpha-divergence minimization is thorough, clear and detailed. The paper provides significant theoretical insight and understanding into alpha-divergence minimization and optimization-based approximate inference in general. My biggest concern about the alpha-divergence framework is whether its theoretical richness and elegance actually translates to practical methods. In other words, I'm not sure that the practical aspects of it are appealing enough to convince practitioners of variational inference to switch to alpha-divergence minimization instead.

The paper refers to [2] and says that those authors proved statistical consistency. However, I am then surprised to see in section 4.3 that non-zero shrinkage is obtained (including for gamma 0) for the very simple case of modelling a N(0,I) distribution with N(0, sigma 2 I). What is going on here?? A failure of consistency would be a serious flaw in the formulation of a statistical learning criterion. Also in sec 3 (Stability and KL regularization) the authors say that at least for learning based on samples (\hat{p}_{theta}) that some regularization wrt the KL divergence is required. This clearly weakens the "purity" of the smoothed Wasserstein objective fn.

The idea of probability generating function achieving similar forms of "conjugacy" as Gaussian random variables and (finite) discrete random variables is very interesting. This enables a fast and exact inference algorithm on Poisson latent variable models. The formulation involving probability generating function does not seem to be constrained to Poisson random variables but all the simulations and real application pertain to Poisson HMM. It is unclear if there's anything special with Poisson that enables better parameter estimates. It is also unclear if there are other Poisson latent variable models other than Poisson HMMs with wide applications.

The authors do a good job of presenting the high-level ideas behind their contribution and presenting the relevant literature in context. The actual contribution is quite technical in nature, but a good amount of effort is taken to walk the reader through it. The authors might also look into approaches like SLIM (Ning and Karypis), which also approach matrix completion tasks using (fairly simple) models that overcome the low-rank assumption of typical matrix completion approaches. Although the paper promises to recover matrix data generated by a quite general class of functions, I struggled to understand which of the operating assumptions (section 2) are actually realistic. In particular, assumption (e) (each entry is observed independently) is certainly violated in the netflix and movielens datasets where the "missing at random" assumption does not hold (as would be the case in any dataset where users self-select what to evaluate; see papers on the "missing not at random" assumption).

This is a well-done paper. It attacks a problem that is worthwhile: how to construct and train a sequence-to-sequence model that can operate on-line instead of waiting for an entire input to be received. It clearly describes an architecture for solving the problem, and walks the reader through the issues in the design of each component in the architecture: next-step prediction, the attention mechanism, and modeling the ends of blocks. It clearly explains the challenges that need to be overcome train the model and perform inference with it, and proposes reasonable approximate algorithms for training and inference. The speech recognition experiments used to demonstrate the utility of the transducer model and to explore design issues such as maintenance of recurrent state across block boundaries, block size, design of the attention mechanism, and depth of the model are reasonable.

The paper is a superbly written account of a simple idea that appears to work very well. The approach can straightforwardly be applied to existing max-likelihood (ML) trained models in order to in principle take into account the task reward during training and is computationally much more efficient than alternative non ML based approaches. This work risks being underappreciated as proposing but a simple addition of artificial structured-label noise, but I think the specific link with structured output task reward is sufficiently original, and the paper also uncovers important theoretical insight by revealing the formal relationship between the proposed reward augmented ML and RL-based regularized expected reward objectives. So while it works surprisingly well, you haven't yet clearly demonstrated empirically that using a truly *task-reward derived* payoff distribution is beneficial. One way to convincingly demonstrate that would be if you did your envisioned BLEU importance reweighted sampling, and were able to show that it improves the BLEU test score over your current simpler edit-distance based label noise.

Contextual MDPs are a specific type of POMDPs with the restriction that the optimal q-function depends only on the most recent observation (instead of the belief state). The authors show that Contextual MDPs are not poly PAC learneable even when either memoryless policies are considered or value function approximation is used. However, when both memoryless policies and value function approximation is used and the transitions are deterministic, then the model is PAC learnable in a polynomial number of episodes (and the complexity is independent of the number of observations). The paper is well written overall. The proofs are quite clear and quite thorough. I am not quite sure that the 16 pages of technical proofs in the appendix are suitable for a conference; the paper may better fit a journal format.

The paper is very clearly written and describes technical concepts in a very comprehensible way. The approach is sound and well motivated and the experimental comparisons with other approaches are fair, though they could have been more extensive in terms of datasets. My greatest concern is about the execution time of the proposed approach, since this is a sequential Monte Carlo method that performs multiple refinement passes for each step of the training process. The authors report convergence curves vs epochs but not vs wall clock time, which should be provided as the main motivation of the paper is to speed up training for this class of generative methods. The experimental section is good in terms of which methods it compares against, but a bit lacking in terms of datasets.

This is an interesting modeling and model comparison paper, providing insights into the processing of uncertainty during learning and decision making. The paper combines advances that could be interesting to both experimental and modeling audiences. However, its clarity should be improved and parameter estimation details explained much better for the paper to be acceptable to NIPS. More specifically: - Why should highly volatile environments have high learning rates (line 2 of page 2)? Couldn't it plausibly lead to excessive weight instability?