Options are generally learned by using an inaccurate environment model (or simulator), which contains uncertain model parameters. While there are several methods to learn options that are robust against the uncertainty of model parameters, these methods only consider either the worst case or the average (ordinary) case for learning options. This limited consideration of the cases often produces options that do not work well in the unconsidered case. In this paper, we propose a conditional value at risk (CVaR)-based method to learn options that work well in both the average and worst cases. We extend the CVaR-based policy gradient method proposed by Chow and Ghavamzadeh (2014) to deal with robust Markov decision processes and then apply the extended method to learning robust options. We conduct experiments to evaluate our method in multi-joint robot control tasks (HopperIceBlock, Half-Cheetah, and Walker2D). Experimental results show that our method produces options that 1) give better worst-case performance than the options learned only to minimize the average-case loss, and 2) give better average-case performance than the options learned only to minimize the worst-case loss.

Normalization layers and activation functions are fundamental components in deep networks and typically co-locate with each other. Here we propose to design them using an automated approach. Instead of designing them separately, we unify them into a single tensor-to-tensor computation graph, and evolve its structure starting from basic mathematical functions. Examples of such mathematical functions are addition, multiplication and statistical moments. The use of low-level mathematical functions, in contrast to the use of high-level modules in mainstream NAS, leads to a highly sparse and large search space which can be challenging for search methods.

This new real-time interpreter can change your language while cloning your voice--sort of. Live voice call translator raises the bar on these devices. Screen is tiny, making typing nearly impossible. Voice cloning feature is hit and miss. Real-time translation gadgets get another upgrade with Vasco's latest, a handheld translator with a feature that is decidedly cool, at least on paper: voice cloning technology.

We answer major comments from each reviewer below; we'll fix the minor ones. REVIEWER 1: "This paper ranks high in novelty...The experimental results are strong, especially on T ext Some important details are unclear . E.g. what is the base distribution for sampling? REVIEWER 2: "Originality: This paper is the first demonstration of flow-based models to discrete data. As such, the work is fairly novel....That being said, the main technical contribution amounts to...on top of the We agree about simplicity being a benefit.

First provide a summary of the paper, and then address the following criteria: Quality, clarity, originality and significance. Summary: The paper introduces an iterative extension of NADE (Neural autoregressive distribution estimator), a generative model that uses a neural network with a variable number of inputs to model each conditional in an autoregressive factorization of a joint distribution. The paper builds up on top of an order-agnostic version of NADE where all dimensions not present in the input are modelled independently by the network at each autoregressive step. The main idea introduced in the paper is using a prediction of the missing inputs at each iteration, starting with the marginal probability distribution over the training data, and the factorial (each dimension is predicted independently conditioned on the input) approximation obtained from NADE in the following iterations. The authors hypothesise that prediction in several steps is easier than in one step.

We thank the reviewers for their insightful comments and constructive feedback. We will answer the major points below and address all remaining ones in the final version. This depends on how C (F) is defined. If it is defined to be the Rademacher complexity, then the former is correct. We'll update with a better one for the final version.

Our methods rely on an "oracle" to solve the NP-hard kidney exchange matching problem.