Recent developments in Transformers have opened new interesting areas of research in partially observable reinforcement learning tasks. Results from late 2019 showed that Transformers are able to outperform LSTMs on both memory intense and reactive tasks. In this work we first partially replicate the results shown in Stabilizing Transformers in RL on both reactive and memory based environments. We then show performance improvement coupled with reduced computation when adding adaptive attention span to this Stable Transformer on a challenging DMLab30 environment.

Visual Question Answering (VQA) systems are tasked with answering natural language questions corresponding to a presented image. Current VQA datasets typically contain questions related to the spatial information of objects, object attributes, or general scene questions. Recently, researchers have recognized the need for improving the balance of such datasets to reduce the system's dependency on memorized linguistic features and statistical biases and to allow for improved visual understanding. However, it is unclear as to whether there are any latent patterns that can be used to quantify and explain these failures. To better quantify our understanding of the performance of VQA models, we use a taxonomy of Knowledge Gaps (KGs) to identify/tag questions with one or more types of KGs. Each KG describes the reasoning abilities needed to arrive at a resolution, and failure to resolve gaps indicate an absence of the required reasoning ability. After identifying KGs for each question, we examine the skew in the distribution of the number of questions for each KG. In order to reduce the skew in the distribution of questions across KGs, we introduce a targeted question generation model. This model allows us to generate new types of questions for an image.

In natural language processing, relation extraction seeks to rationally understand unstructured text. Here, we propose a novel SpanBERT-based graph convolutional network (DG-SpanBERT) that extracts semantic features from a raw sentence using the pre-trained language model SpanBERT and a graph convolutional network to pool latent features. Our DG-SpanBERT model inherits the advantage of SpanBERT on learning rich lexical features from large-scale corpus. It also has the ability to capture long-range relations between entities due to the usage of GCN on dependency tree. The experimental results show that our model outperforms other existing dependency-based and sequence-based models and achieves a state-of-the-art performance on the TACRED dataset.

Analogy is core to human cognition. It allows us to solve problems based on prior experience, it governs the way we conceptualize new information, and it even influences our visual perception. The importance of analogy to humans has made it an active area of research in the broader field of artificial intelligence, resulting in data-efficient models that learn and reason in human-like ways. While analogy and deep learning have generally been considered independently of one another, the integration of the two lines of research seems like a promising step towards more robust and efficient learning techniques. As part of the first steps towards such an integration, we introduce the Analogical Matching Network; a neural architecture that learns to produce analogies between structured, symbolic representations that are largely consistent with the principles of Structure-Mapping Theory.

Dialogue state tracking (DST) aims at estimating the current dialogue state given all the preceding conversation. For multi-domain DST, the data sparsity problem is a major obstacle due to increased numbers of state candidates and dialogue lengths. To encode the dialogue context efficiently, we propose to utilize the previous dialogue state (predicted) and the current dialogue utterance as the input for DST. To consider relations among different domain-slots, the schema graph involving prior knowledge is exploited. In this paper, a novel context and schema fusion network is proposed to encode the dialogue context and schema graph by using internal and external attention mechanisms. Experiment results show that our approach can obtain new state-of-the-art performance of the open-vocabulary DST on both MultiWOZ 2.0 and MultiWOZ 2.1 benchmarks.

Continual Learning has been often framed as the problem of training a model in a sequence of tasks. In this regard, Neural Networks have been attested to forget the solutions to previous task as they learn new ones. Yet, modelling human life-long learning does not necessarily require any crisp notion of tasks. In this work, we propose a benchmark based on language modelling in a multilingual and multidomain setting that prescinds of any explicit delimitation of training examples into distinct tasks, and propose metrics to study continual learning and catastrophic forgetting in this setting. Then, we introduce a simple Product of Experts learning system that performs strongly on this problem while displaying interesting properties, and investigate its merits for avoiding forgetting.

The demand for more transparency of decision-making processes of deep reinforcement learning agents is greater than ever, due to their increased use in safety critical and ethically challenging domains such as autonomous driving. In this empirical study, we address this lack of transparency following an idea that is inspired by research in the field of neuroscience. We characterize the learned representations of an agent's policy network through its activation space and perform partial network ablations to compare the representations of the healthy and the intentionally damaged networks. We show that the healthy agent's behavior is characterized by a distinct correlation pattern between the network's layer activation and the performed actions during an episode and that network ablations, which cause a strong change of this pattern, lead to the agent failing its trained control task. Furthermore, the learned representation of the healthy agent is characterized by a distinct pattern in its activation space reflecting its different behavioral stages during an episode, which again, when distorted by network ablations, leads to the agent failing its trained control task. Concludingly, we argue in favor of a new perspective on artificial neural networks as objects of empirical investigations, just as biological neural systems in neuroscientific studies, paving the way towards a new standard of scientific falsifiability with respect to research on transparency and interpretability of artificial neural networks.

We present Catalyst.RL, an open-source PyTorch framework for reproducible and sample efficient reinforcement learning (RL) research. Main features of Catalyst.RL include large-scale asynchronous distributed training, efficient implementations of various RL algorithms and auxiliary tricks, such as n-step returns, value distributions, hyperbolic reinforcement learning, etc. To demonstrate the effectiveness of Catalyst.RL, we applied it to a physics-based reinforcement learning challenge "NeurIPS 2019: Learn to Move -- Walk Around" with the objective to build a locomotion controller for a human musculoskeletal model. The environment is computationally expensive, has a high-dimensional continuous action space and is stochastic. Our team took the 2nd place, capitalizing on the ability of Catalyst.RL to train high-quality and sample-efficient RL agents in only a few hours of training time. The implementation along with experiments is open-sourced so results can be reproduced and novel ideas tried out.

An artificial intelligence that can predict how a piece of glass responds to heat and pressure could one day also be used to model traffic flow. While most solid materials have a regular atomic structure, the atoms in glass have a more irregular arrangement, resembling a liquid that has been frozen in place. Physicists have long wanted to know more about this "glass transition". "Given that glasses are everywhere – from windows to your phone screen – it's odd that we don't really understand its structure and dynamism," says Victor Bapst at AI firm DeepMind.

Edge computing moves workloads from centralized locations to remote locations and it can provide faster response from AI applications. Edge computing devices are getting deployed increasingly for monitoring and control of real world processes like people tracking, vehicle recognition, pollution monitoring etc. The data collected at the devices gets transported to centralized cloud servers over data pipelines and are used to train machine learning models. Training models needs lot of computational power and the current strategy is to train centrally and deploy on edge devices for inference. Already deep learning models are being used at the edge for critical problems like face recognition and surveillance.