Training task-completion dialogue agents with reinforcement learning usually requires a large number of real user experiences. The Dyna-Q algorithm extends Q-learning by integrating a world model, and thus can effectively boost training efficiency using simulated experiences generated by the world model. The effectiveness of Dyna-Q, however, depends on the quality of the world model - or implicitly, the pre-specified ratio of real vs. simulated experiences used for Q-learning. To this end, we extend the recently proposed Deep Dyna-Q (DDQ) framework by integrating a switcher that automatically determines whether to use a real or simulated experience for Q-learning. Furthermore, we explore the use of active learning for improving sample efficiency, by encouraging the world model to generate simulated experiences in the state-action space where the agent has not (fully) explored. Our results show that by combining switcher and active learning, the new framework named as Switch-based Active Deep Dyna-Q (Switch-DDQ), leads to significant improvement over DDQ and Q-learning baselines in both simulation and human evaluations.

The ability to sequence genomes quickly has provided scientists with reams of data, but understanding how evolution has shaped humans is still a difficult task.Credit: Guy Tear/Wellcome Coll./CC Pinpointing where and how the human genome is evolving can be like hunting for a needle in a haystack. Each person's genome contains three billion building blocks called nucleotides, and researchers must compile data from thousands of people to discover patterns that signal how genes have been shaped by evolutionary pressures. To find these patterns, a growing number of geneticists are turning to a form of machine learning called deep learning. Proponents of the approach say that deep-learning algorithms incorporate fewer explicit assumptions about what the genetic signatures of natural selection should look like than do conventional statistical methods.

Building a cool machine learning project is one thing, but at the end of the day, you want other people to be able to see your hard work. Sure, you could put the whole project on GitHub, but how are your grandparents supposed to figure that out? No, what we want is to deploy our deep learning model as a web application accessible to anyone in the world. In this article, we'll see how to write a web application that takes a trained Keras recurrent neural network and allows users to generate new patent abstracts. This project builds on work from the Recurrent Neural Networks by Example article, but knowing how to create the RNN isn't necessary.

Building a cool machine learning project is one thing, but at the end of the day, you want other people to be able to see your hard work. Sure, you could put the whole project on GitHub, but how are your grandparents supposed to figure that out? No, what we want is to deploy our deep learning model as a web application accessible to anyone in the world. In this article, we'll see how to write a web application that takes a trained Keras recurrent neural network and allows users to generate new patent abstracts. This project builds on work from the Recurrent Neural Networks by Example article, but knowing how to create the RNN isn't necessary.

When U.N. member states unanimously adopted the 2030 Agenda in 2015, the narrative around global development embraced a new paradigm of sustainability and inclusion--of planetary stewardship alongside economic progress, and inclusive distribution of income. This comprehensive agenda--merging social, economic and environmental dimensions of sustainability--is not supported by current modes of data collection and data analysis, so the report of the High-Level Panel on the post-2015 development agenda called for a "data revolution" to empower people through access to information.1 Today, a central development problem is that high-quality, timely, accessible data are absent in most poor countries, where development needs are greatest. In a world of unequal distributions of income and wealth across space, age and class, gender and ethnic pay gaps, and environmental risks, data that provide only national averages conceal more than they reveal. This paper argues that spatial disaggregation and timeliness could permit a process of evidence-based policy making that monitors outcomes and adjusts actions in a feedback loop that can accelerate development through learning. Big data and artificial intelligence are key elements in such a process. Emerging technologies could lead to the next quantum leap in (i) how data is collected; (ii) how data is analyzed; and (iii) how analysis is used for policymaking and the achievement of better results. Big data platforms expand the toolkit for acquiring real-time information at a granular level, while machine learning permits pattern recognition across multiple layers of input. Together, these advances could make data more accessible, scalable, and finely tuned. In turn, the availability of real-time information can shorten the feedback loop between results monitoring, learning, and policy formulation or investment, accelerating the speed and scale at which development actors can implement change.

If the robots are going to inherit from humans not just our minds, but our filthy little souls too, then we're all probably doomed. Ongoing research from Google's DeepMind AI project reportedly indicates that game-playing artificial intelligence is developing a nasty cheating habit, if cheating's what it takes to finish atop the leaderboard. Setting an AI loose on a video game with the simple instruction to rack up as many points as possible, it turns out, is basically an invitation to have it find and exploit every loophole it can. Even if it means hurling yourself over a ledge, kamikaze-style, in a self-sacrificing loop, or pausing the game just at the moment your victory's in doubt, AI is learning how to win at all costs – or at least make sure that nobody else does. One AI that was learning how to play Q*bert, reports Kotaku, "Even took to killing itself to boost its score. After discovering a pattern of movement by which it could get enemies to follow it off a cliff in order to gain more points and an extra life, it continued to do just that for the rest of the session."

In chemistry, deep neural network models have been increasingly utilized in a variety of applications such as molecular property predictions, novel molecule designs, and planning chemical reactions. Despite the rapid increase in the use of state-of-the-art models and algorithms, deep neural network models often produce poor predictions in real applications because model performance is highly dependent on the quality of training data. In the field of molecular analysis, data are mostly obtained from either complicated chemical experiments or approximate mathematical equations, and then quality of data may be questioned. In this paper, we quantify uncertainties of prediction using Bayesian neural networks in molecular property predictions. We estimate both model-driven and data-driven uncertainties, demonstrating the usefulness of uncertainty quantification as both a quality checker and a confidence indicator with the three experiments. Our results manifest that uncertainty quantification is necessary for more reliable molecular applications and Bayesian neural network models can be a practical approach.

Natural spatiotemporal processes can be highly non-stationary in many ways, e.g. the low-level non-stationarity such as spatial correlations or temporal dependencies of local pixel values; and the high-level variations such as the accumulation, deformation or dissipation of radar echoes in precipitation forecasting. From Cramer's Decomposition, any non-stationary process can be decomposed into deterministic, time-variant polynomials, plus a zero-mean stochastic term. By applying differencing operations appropriately, we may turn time-variant polynomials into a constant, making the deterministic component predictable. However, most previous recurrent neural networks for spatiotemporal prediction do not use the differential signals effectively, and their relatively simple state transition functions prevent them from learning too complicated variations in spacetime. We propose the Memory In Memory (MIM) networks and corresponding recurrent blocks for this purpose. The MIM blocks exploit the differential signals between adjacent recurrent states to model the non-stationary and approximately stationary properties in spatiotemporal dynamics with two cascaded, self-renewed memory modules. By stacking multiple MIM blocks, we could potentially handle higher-order non-stationarity. The MIM networks achieve the state-of-the-art results on three spatiotemporal prediction tasks across both synthetic and real-world datasets. We believe that the general idea of this work can be potentially applied to other time-series forecasting tasks.

Deep neural networks (DNNs) have set benchmarks on a wide array of supervised learning tasks. Trained DNNs, however, often lack robustness to minor adversarial perturbations to the input, which undermines their true practicality. Recent works have increased the robustness of DNNs by fitting networks using adversarially-perturbed training samples, but the improved performance can still be far below the performance seen in non-adversarial settings. A significant portion of this gap can be attributed to the decrease in generalization performance due to adversarial training. In this work, we extend the notion of margin loss to adversarial settings and bound the generalization error for DNNs trained under several well-known gradient-based attack schemes, motivating an effective regularization scheme based on spectral normalization of the DNN's weight matrices. We also provide a computationally-efficient method for normalizing the spectral norm of convolutional layers with arbitrary stride and padding schemes in deep convolutional networks. We evaluate the power of spectral normalization extensively on combinations of datasets, network architectures, and adversarial training schemes. The code is available at https://github.com/jessemzhang/dl_spectral_normalization.

Machine learning is increasingly targeting areas where input data cannot be accurately described by a single vector, but can be modeled instead using the more flexible concept of random vectors, namely probability measures or more simply point clouds of varying cardinality. Using deep architectures on measures poses, however, many challenging issues. Indeed, deep architectures are originally designed to handle fixedlength vectors, or, using recursive mechanisms, ordered sequences thereof. In sharp contrast, measures describe a varying number of weighted observations with no particular order. We propose in this work a deep framework designed to handle crucial aspects of measures, namely permutation invariances, variations in weights and cardinality. Architectures derived from this pipeline can (i) map measures to measures - using the concept of push-forward operators; (ii) bridge the gap between measures and Euclidean spaces - through integration steps. This allows to design discriminative networks (to classify or reduce the dimensionality of input measures), generative architectures (to synthesize measures) and recurrent pipelines (to predict measure dynamics). We provide a theoretical analysis of these building blocks, review our architectures' approximation abilities and robustness w.r.t. perturbation, and try them on various discriminative and generative tasks.