The field of meta-learning, or learning-to-learn, has seen a dramatic rise in interest in recent years. Contrary to conventional approaches to AI where a given task is solved from scratch using a fixed learning algorithm, meta-learning aims to improve the learning algorithm itself, given the experience of multiple learning episodes. This paradigm provides an opportunity to tackle many of the conventional challenges of deep learning, including data and computation bottlenecks, as well as the fundamental issue of generalization. In this survey we describe the contemporary meta-learning landscape. We first discuss definitions of meta-learning and position it with respect to related fields, such as transfer learning, multi-task learning, and hyperparameter optimization. We then propose a new taxonomy that provides a more comprehensive breakdown of the space of meta-learning methods today. We survey promising applications and successes of meta-learning including few-shot learning, reinforcement learning and architecture search. Finally, we discuss outstanding challenges and promising areas for future research.

Demonstration is an appealing way for humans to provide assistance to reinforcement-learning agents. Most approaches in this area view demonstrations primarily as sources of behavioral bias. But in sparse-reward tasks, humans seem to treat demonstrations more as sources of causal knowledge. This paper proposes a framework for agents that benefit from demonstration in this human-inspired way. In this framework, agents develop causal models through observation, and reason from this knowledge to decompose tasks for effective reinforcement learning. Experimental results show that a basic implementation of Reasoning from Demonstration (RfD) is effective in a range of sparse-reward tasks.

"Entities should not be multiplied unnecessarily" To incentivize the scientific community to focus on AGI, Marcus Hutter, one of the most prominent researchers of our generation, has renewed his decade-old prize by ten folds to half a million euros (500,000 €). The Hutter prize, named after Marcus Hutter, is given to those who can successfully create new benchmarks for lossless data compression. The data here is a dataset based on Wikipedia. Marcus Hutter, who now works at DeepMind as a senior research scientist, is famous for his work on reinforcement learning along with Juergen Schmidhuber. Dr Hutter proposed AIXI in 2000, which is a reinforcement learning agent that works in line with Occam's razor and sequential decision theory.

While deep neural networks (DNNs) and Gaussian Processes (GPs) are both popularly utilized to solve problems in reinforcement learning, both approaches feature undesirable drawbacks for challenging problems. DNNs learn complex nonlinear embeddings, but do not naturally quantify uncertainty and are often data-inefficient to train. GPs infer posterior distributions over functions, but popular kernels exhibit limited expressivity on complex and high-dimensional data. Fortunately, recently discovered conjugate and neural tangent kernel functions encode the behavior of overparameterized neural networks in the kernel domain. We demonstrate that these kernels can be efficiently applied to regression and reinforcement learning problems by analyzing a baseline case study. We apply GPs with neural network dual kernels to solve reinforcement learning tasks for the first time. We demonstrate, using the well-understood mountain-car problem, that GPs empowered with dual kernels perform at least as well as those using the conventional radial basis function kernel. We conjecture that by inheriting the probabilistic rigor of GPs and the powerful embedding properties of DNNs, GPs using NN dual kernels will empower future reinforcement learning models on difficult domains.

This work presents a modular and hierarchical approach to learn policies for exploring 3D environments, called `Active Neural SLAM'. Our approach leverages the strengths of both classical and learning-based methods, by using analytical path planners with learned SLAM module, and global and local policies. The use of learning provides flexibility with respect to input modalities (in the SLAM module), leverages structural regularities of the world (in global policies), and provides robustness to errors in state estimation (in local policies). Such use of learning within each module retains its benefits, while at the same time, hierarchical decomposition and modular training allow us to sidestep the high sample complexities associated with training end-to-end policies. Our experiments in visually and physically realistic simulated 3D environments demonstrate the effectiveness of our approach over past learning and geometry-based approaches. The proposed model can also be easily transferred to the PointGoal task and was the winning entry of the CVPR 2019 Habitat PointGoal Navigation Challenge.

Reinforcement learning agents learn by encouraging behaviours which maximize their total reward, usually provided by the environment. In many environments, however, the reward is provided after a series of actions rather than each single action, causing the agent to experience ambiguity in terms of whether those actions are effective, an issue called the credit assignment problem. In this paper, we propose two strategies, inspired by behavioural psychology, to estimate a more informative reward value for actions with no reward. The first strategy, called self-punishment, discourages the agent to avoid making mistakes, i.e., actions which lead to a terminal state. The second strategy, called the rewards backfill, backpropagates the rewards between two rewarded actions. We prove that, under certain assumptions, these two strategies maintain the order of the policies in the space of all possible policies in terms of their total reward, and, by extension, maintain the optimal policy. We incorporated these two strategies into three popular deep reinforcement learning approaches and evaluated the results on thirty Atari games. After parameter tuning, our results indicate that the proposed strategies improve the tested methods in over 65 percent of tested games by up to over 25 times performance improvement.

Reinforcement learning is nowadays a popular framework for solving different decision making problems in automated driving. However, there are still some remaining crucial challenges that need to be addressed for providing more reliable policies. In this paper, we propose a generic risk-aware DQN approach in order to learn high level actions for driving through unsignalized occluded intersections. The proposed state representation provides lane based information which allows to be used for multi-lane scenarios. Moreover, we propose a risk based reward function which punishes risky situations instead of only collision failures. Such rewarding approach helps to incorporate risk prediction into our deep Q network and learn more reliable policies which are safer in challenging situations. The efficiency of the proposed approach is compared with a DQN learned with conventional collision based rewarding scheme and also with a rule-based intersection navigation policy. Evaluation results show that the proposed approach outperforms both of these methods. It provides safer actions than collision-aware DQN approach and is less overcautious than the rule-based policy.

This paper considers policy search in continuous state-action reinforcement learning problems. Typically, one computes search directions using a classic expression for the policy gradient called the Policy Gradient Theorem, which decomposes the gradient of the value function into two factors: the score function and the Q-function. This paper presents four results:(i) an alternative policy gradient theorem using weak (measure-valued) derivatives instead of score-function is established; (ii) the stochastic gradient estimates thus derived are shown to be unbiased and to yield algorithms that converge almost surely to stationary points of the non-convex value function of the reinforcement learning problem; (iii) the sample complexity of the algorithm is derived and is shown to be $O(1/\sqrt(k))$; (iv) finally, the expected variance of the gradient estimates obtained using weak derivatives is shown to be lower than those obtained using the popular score-function approach. Experiments on OpenAI gym pendulum environment show superior performance of the proposed algorithm.

In reinforcement learning (RL), dealing with non-stationarity is a challenging issue. However, some domains such as traffic optimization are inherently non-stationary. Causes for and effects of this are manifold. In particular, when dealing with traffic signal controls, addressing non-stationarity is key since traffic conditions change over time and as a function of traffic control decisions taken in other parts of a network. In this paper we analyze the effects that different sources of non-stationarity have in a network of traffic signals, in which each signal is modeled as a learning agent. More precisely, we study both the effects of changing the \textit{context} in which an agent learns (e.g., a change in flow rates experienced by it), as well as the effects of reducing agent observability of the true environment state. Partial observability may cause distinct states (in which distinct actions are optimal) to be seen as the same by the traffic signal agents. This, in turn, may lead to sub-optimal performance. We show that the lack of suitable sensors to provide a representative observation of the real state seems to affect the performance more drastically than the changes to the underlying traffic patterns.

We undertake a precise study of the asymptotic and non-asymptotic properties of stochastic approximation procedures with Polyak-Ruppert averaging for solving a linear system $\bar{A} \theta = \bar{b}$. When the matrix $\bar{A}$ is Hurwitz, we prove a central limit theorem (CLT) for the averaged iterates with fixed step size and number of iterations going to infinity. The CLT characterizes the exact asymptotic covariance matrix, which is the sum of the classical Polyak-Ruppert covariance and a correction term that scales with the step size. Under assumptions on the tail of the noise distribution, we prove a non-asymptotic concentration inequality whose main term matches the covariance in CLT in any direction, up to universal constants. When the matrix $\bar{A}$ is not Hurwitz but only has non-negative real parts in its eigenvalues, we prove that the averaged LSA procedure actually achieves an $O(1/T)$ rate in mean-squared error. Our results provide a more refined understanding of linear stochastic approximation in both the asymptotic and non-asymptotic settings. We also show various applications of the main results, including the study of momentum-based stochastic gradient methods as well as temporal difference algorithms in reinforcement learning.