Supporting state-of-the-art AI research requires balancing rapid prototyping, ease of use, and quick iteration, with the ability to deploy experiments at a scale traditionally associated with production systems.Deep learning frameworks such as TensorFlow, PyTorch and JAX allow users to transparently make use of accelerators, such as TPUs and GPUs, to offload the more computationally intensive parts of training and inference in modern deep learning systems. Popular training pipelines that use these frameworks for deep learning typically focus on (un-)supervised learning. How to best train reinforcement learning (RL) agents at scale is still an active research area. In this report we argue that TPUs are particularly well suited for training RL agents in a scalable, efficient and reproducible way. Specifically we describe two architectures designed to make the best use of the resources available on a TPU Pod (a special configuration in a Google data center that features multiple TPU devices connected to each other by extremely low latency communication channels).

A major driver behind the success of modern machine learning algorithms has been their ability to process ever-larger amounts of data. As a result, the use of distributed systems in both research and production has become increasingly prevalent as a means to scale to this growing data. At the same time, however, distributing the learning process can drastically complicate the implementation of even simple algorithms. This is especially problematic as many machine learning practitioners are not well-versed in the design of distributed systems, let alone those that have complicated communication topologies. In this work we introduce Launchpad, a programming model that simplifies the process of defining and launching distributed systems that is specifically tailored towards a machine learning audience. We describe our framework, its design philosophy and implementation, and give a number of examples of common learning algorithms whose designs are greatly simplified by this approach.

A central component of training in Reinforcement Learning (RL) is Experience: the data used for training. The mechanisms used to generate and consume this data have an important effect on the performance of RL algorithms. In this paper, we introduce Reverb: an efficient, extensible, and easy to use system designed specifically for experience replay in RL. Reverb is designed to work efficiently in distributed configurations with up to thousands of concurrent clients. The flexible API provides users with the tools to easily and accurately configure the replay buffer. It includes strategies for selecting and removing elements from the buffer, as well as options for controlling the ratio between sampled and inserted elements. This paper presents the core design of Reverb, gives examples of how it can be applied, and provides empirical results of Reverb's performance characteristics.

The objective of a reinforcement learning agent is to behave so as to maximise the sum of a suitable scalar function of state: the reward. These rewards are typically given and immutable. In this paper, we instead consider the proposition that the reward function itself can be a good locus of learned knowledge. To investigate this, we propose a scalable meta-gradient framework for learning useful intrinsic reward functions across multiple lifetimes of experience. Through several proof-of-concept experiments, we show that it is feasible to learn and capture knowledge about long-term exploration and exploitation into a reward function. Furthermore, we show that unlike policy transfer methods that capture how the agent should behave, the learned reward functions can generalise to other kinds of agents and to changes in the dynamics of the environment by capturing what the agent should strive to do.

Reinforcement learning agents can include different components, such as policies, value functions, state representations, and environment models. Any or all of these can be the loci of knowledge, i.e., structures where knowledge, whether given or learned, can be deposited and reused. Regardless of its composition, the objective of an agent is behave so as to maximise the sum of a suitable scalar function of state: the reward. As far as the learning algorithm is concerned, these rewards are typically given and immutable. In this paper we instead consider the proposition that the reward function itself may be a good locus of knowledge. This is consistent with a common use, in the literature, of hand-designed intrinsic rewards to improve the learning dynamics of an agent. We adopt the multi-lifetime setting of the Optimal Rewards Framework, and propose to meta-learn an intrinsic reward function from experience that allows agents to maximise their extrinsic rewards accumulated until the end of their lifetimes. Rewards as a locus of knowledge provide guidance on “what” the agent should strive to do rather than on knowledge of “how” the agent should behave that is more directly captured in policies or value functions for example. Thus, our focus here is on demonstrating the following: (1) that it is feasible to meta-learn good reward functions, (2) to show that the learned reward functions can capture interesting kinds of “what” knowledge, and (3) that because of the indirectness of this form of knowledge the learned reward functions can generalise to other kinds of agents and to changes in the dynamics of the environment. Reinforcement learning agents can store knowledge in their policies, value functions, state representations, and models of the environment dynamics. These components can be the loci of knowledge in the sense that they are structures in which knowledge, either learned from experience by the agent’s algorithm or given by the agent-designer, can be deposited and reused. The objective of the agent is defined by a reward function, and the goal is to learn to act so as to optimise cumulative rewards. In this paper we consider the proposition that the reward function itself is a good locus of knowledge. This is unusual in that most prior work treats the reward as given and immutable, at least as far as the learning algorithm is concerned. At the same time, especially in challenging reinforcement-learning problems, agent designers do find it convenient to modify the reward function given to the agent to facilitate learning. It is therefore useful to distinguish between two kinds of reward functions (Singh et al., 2010): extrinsic rewards define the task and capture the designer’s preferences over agent behaviour, whereas intrinsic rewards serve as helpful signals to improve the learning dynamics of the agent. Intrinsic rewards are typically hand-designed and then often added to the immutable extrinsic rewards to form the reward optimised by the agent. Most existing work on intrinsic rewards falls into two broad categories: task-dependent and taskindependent. Both are typically designed by hand. Hand-designing task-dependent rewards can be fraught with difficulty as even minor misalignment between the actual reward and the intended bias can lead to unintended and sometimes catastrophic consequences (Clark & Amodei, 2016). Task-independent intrinsic rewards are also typically hand-designed, often based on an intuitive understanding of animal/human behaviour or on heuristics on desired exploratory behaviour. It can, however, be hard to match such task-independent intrinsic rewards to the specific learning dynamics induced by the interaction between agent and environment. The motivation for this paper is our interest in the comparatively under-explored possibility of learned (not hand-designed) taskdependent intrinsic rewards (see Zheng et al., 2018, for previous work). We emphasise that it is not our objective to show that rewards are a better locus of learned knowledge than others; the best locus likely depends on the kind of knowledge that is most useful in a given

Many real-world applications require artificial agents to compete and coordinate with other agents in complex environments. As a stepping stone to this goal, the domain of StarCraft has emerged as an important challenge for artificial intelligence research, owing to its iconic and enduring status among the most difficult professional esports and its relevance to the real world in terms of its raw complexity and multi-agent challenges. Over the course of a decade and numerous competitions(1-3), the strongest agents have simplified important aspects of the game, utilized superhuman capabilities, or employed hand-crafted sub-systems(4). Despite these advantages, no previous agent has come close to matching the overall skill of top StarCraft players. We chose to address the challenge of StarCraft using general-purpose learning methods that are in principle applicable to other complex domains: a multi-agent reinforcement learning algorithm that uses data from both human and agent games within a diverse league of continually adapting strategies and counter-strategies, each represented by deep neural networks(5,6). We evaluated our agent, AlphaStar, in the full game of StarCraft II, through a series of online games against human players. AlphaStar was rated at Grandmaster level for all three StarCraft races and above 99.8% of officially ranked human players.

Differentiation alters molecular properties of stem and progenitor cells, leading to changes in their shape and movement characteristics. We present a deep neural network that prospectively predicts lineage choice in differentiating primary hematopoietic progenitors using image patches from brightfield microscopy and cellular movement. Surprisingly, lineage choice can be detected up to three generations before conventional molecular markers are observable. Our approach allows identification of cells with differentially expressed lineage-specifying genes without molecular labeling.

RNA-sequencing has revolutionized biomedical research and, in particular, our ability to study gene alternative splicing. The problem has important implications for human health, as alternative splicing may be involved in malfunctions at the cellular level and multiple diseases. However, the high-dimensional nature of the data and the existence of experimental biases pose serious data analysis challenges. We find that the standard data summaries used to study alternative splicing are severely limited, as they ignore a substantial amount of valuable information. Current data analysis methods are based on such summaries and are hence sub-optimal. Further, they have limited flexibility in accounting for technical biases. We propose novel data summaries and a Bayesian modeling framework that overcome these limitations and determine biases in a non-parametric, highly flexible manner. These summaries adapt naturally to the rapid improvements in sequencing technology. We provide efficient point estimates and uncertainty assessments. The approach allows to study alternative splicing patterns for individual samples and can also be the basis for downstream analyses. We found a several fold improvement in estimation mean square error compared popular approaches in simulations, and substantially higher consistency between replicates in experimental data. Our findings indicate the need for adjusting the routine summarization and analysis of alternative splicing RNA-seq studies. We provide a software implementation in the R package casper.