Unlocking Pixels for Reinforcement Learning via Implicit Attention

There has recently been significant interest in training reinforcement learning (RL) agents in vision-based environments. This poses many challenges, such as high dimensionality and potential for observational overfitting through spurious correlations. A promising approach to solve both of these problems is a self-attention bottleneck, which provides a simple and effective framework for learning high performing policies, even in the presence of distractions. However, due to poor scalability of attention architectures, these methods do not scale beyond low resolution visual inputs, using large patches (thus small attention matrices). In this paper we make use of new efficient attention algorithms, recently shown to be highly effective for Transformers, and demonstrate that these new techniques can be applied in the RL setting. This allows our attention-based controllers to scale to larger visual inputs, and facilitate the use of smaller patches, even individual pixels, improving generalization. In addition, we propose a new efficient algorithm approximating softmax attention with what we call hybrid random features, leveraging the theory of angular kernels. We show theoretically and empirically that hybrid random features is a promising approach when using attention for vision-based RL.

Deep Reinforcement Learning with Dynamic Optimism

In recent years, deep off-policy actor-critic algorithms have become a dominant approach to reinforcement learning for continuous control. This comes after a series of breakthroughs to address function approximation errors, which previously led to poor performance. These insights encourage the use of pessimistic value updates. However, this discourages exploration and runs counter to theoretical support for the efficacy of optimism in the face of uncertainty. So which approach is best? In this work, we show that the optimal degree of optimism can vary both across tasks and over the course of learning. Inspired by this insight, we introduce a novel deep actor-critic algorithm, Dynamic Optimistic and Pessimistic Estimation (DOPE) to switch between optimistic and pessimistic value learning online by formulating the selection as a multi-arm bandit problem. We show in a series of challenging continuous control tasks that DOPE outperforms existing state-of-the-art methods, which rely on a fixed degree of optimism. Since our changes are simple to implement, we believe these insights can be extended to a number of off-policy algorithms.

ES-ENAS: Combining Evolution Strategies with Neural Architecture Search at No Extra Cost for Reinforcement Learning

We introduce ES-ENAS, a simple neural architecture search (NAS) algorithm for the purpose of reinforcement learning (RL) policy design, by combining Evolutionary Strategies (ES) and Efficient NAS (ENAS) in a highly scalable and intuitive way. Our main insight is noticing that ES is already a distributed blackbox algorithm, and thus we may simply insert a model controller from ENAS into the central aggregator in ES and obtain weight sharing properties for free. By doing so, we bridge the gap from NAS research in supervised learning settings to the reinforcement learning scenario through this relatively simple marriage between two different lines of research, and are one of the first to apply controller-based NAS techniques to RL. We demonstrate the utility of our method by training combinatorial neural network architectures for RL problems in continuous control, via edge pruning and weight sharing. We also incorporate a wide variety of popular techniques from modern NAS literature, including multiobjective optimization and varying controller methods, to showcase their promise in the RL field and discuss possible extensions. We achieve >90% network compression for multiple tasks, which may be special interest in mobile robotics with limited storage and computational resources.

Ridge Rider: Finding Diverse Solutions by Following Eigenvectors of the Hessian

Over the last decade, a single algorithm has changed many facets of our lives - Stochastic Gradient Descent (SGD). In the era of ever decreasing loss functions, SGD and its various offspring have become the go-to optimization tool in machine learning and are a key component of the success of deep neural networks (DNNs). While SGD is guaranteed to converge to a local optimum (under loose assumptions), in some cases it may matter which local optimum is found, and this is often context-dependent. Examples frequently arise in machine learning, from shape-versus-texture-features to ensemble methods and zero-shot coordination. In these settings, there are desired solutions which SGD on 'standard' loss functions will not find, since it instead converges to the 'easy' solutions. In this paper, we present a different approach. Rather than following the gradient, which corresponds to a locally greedy direction, we instead follow the eigenvectors of the Hessian, which we call "ridges". By iteratively following and branching amongst the ridges, we effectively span the loss surface to find qualitatively different solutions. We show both theoretically and experimentally that our method, called Ridge Rider (RR), offers a promising direction for a variety of challenging problems.

On Optimism in Model-Based Reinforcement Learning

The principle of optimism in the face of uncertainty is prevalent throughout sequential decision making problems such as multi-armed bandits and reinforcement learning (RL), often coming with strong theoretical guarantees. However, it remains a challenge to scale these approaches to the deep RL paradigm, which has achieved a great deal of attention in recent years. In this paper, we introduce a tractable approach to optimism via noise augmented Markov Decision Processes (MDPs), which we show can obtain a competitive regret bound: $\tilde{\mathcal{O}}( |\mathcal{S}|H\sqrt{|\mathcal{S}||\mathcal{A}| T } )$ when augmenting using Gaussian noise, where $T$ is the total number of environment steps. This tractability allows us to apply our approach to the deep RL setting, where we rigorously evaluate the key factors for success of optimistic model-based RL algorithms, bridging the gap between theory and practice.

Stochastic Flows and Geometric Optimization on the Orthogonal Group

We present a new class of stochastic, geometrically-driven optimization algorithms on the orthogonal group $O(d)$ and naturally reductive homogeneous manifolds obtained from the action of the rotation group $SO(d)$. We theoretically and experimentally demonstrate that our methods can be applied in various fields of machine learning including deep, convolutional and recurrent neural networks, reinforcement learning, normalizing flows and metric learning. We show an intriguing connection between efficient stochastic optimization on the orthogonal group and graph theory (e.g. matching problem, partition functions over graphs, graph-coloring). We leverage the theory of Lie groups and provide theoretical results for the designed class of algorithms. We demonstrate broad applicability of our methods by showing strong performance on the seemingly unrelated tasks of learning world models to obtain stable policies for the most difficult $\mathrm{Humanoid}$ agent from $\mathrm{OpenAI}$ $\mathrm{Gym}$ and improving convolutional neural networks.

Ready Policy One: World Building Through Active Learning

Model-Based Reinforcement Learning (MBRL) offers a promising direction for sample efficient learning, often achieving state of the art results for continuous control tasks. However, many existing MBRL methods rely on combining greedy policies with exploration heuristics, and even those which utilize principled exploration bonuses construct dual objectives in an ad hoc fashion. In this paper we introduce Ready Policy One (RP1), a framework that views MBRL as an active learning problem, where we aim to improve the world model in the fewest samples possible. RP1 achieves this by utilizing a hybrid objective function, which crucially adapts during optimization, allowing the algorithm to trade off reward v.s. exploration at different stages of learning. In addition, we introduce a principled mechanism to terminate sample collection once we have a rich enough trajectory batch to improve the model. We rigorously evaluate our method on a variety of continuous control tasks, and demonstrate statistically significant gains over existing approaches.

One-Shot Bayes Opt with Probabilistic Population Based Training

Selecting optimal hyperparameters is a key challenge in machine learning. An exciting recent result showed it is possible to learn high-performing hyperparameter schedules on the fly in a single training run through methods inspired by Evolutionary Algorithms. These approaches have been shown to increase performance across a wide variety of machine learning tasks, ranging from supervised (SL) to reinforcement learning (RL). However, since they remain primarily evolutionary, they act in a greedy fashion, thus require a combination of vast computational resources and carefully selected meta-parameters to effectively explore the hyperparameter space. To address these shortcomings we look to Bayesian Optimization (BO), where a Gaussian Process surrogate model is combined with an acquisition function to produce a principled mechanism to trade off exploration vs exploitation. Our approach, which we call Probabilistic Population-Based Training ($\mathrm{P2BT}$), is able to transfer sample efficiency of BO to the online setting, making it possible to achieve these traits in a single training run. We show that $\mathrm{P2BT}$ is able to achieve high performance with only a small population size, making it useful for all researchers regardless of their computational resources.

Effective Diversity in Population-Based Reinforcement Learning

Maintaining a population of solutions has been shown to increase exploration in reinforcement learning, typically attributed to the greater diversity of behaviors considered. One such class of methods, novelty search, considers further boosting diversity across agents via a multi-objective optimization formulation. Despite the intuitive appeal, these mechanisms have several shortcomings. First, they make use of mean field updates, which induce cycling behaviors. Second, they often rely on handcrafted behavior characterizations, which require domain knowledge. Furthermore, boosting diversity often has a detrimental impact on optimizing already fruitful behaviors for rewards. Setting the relative importance of novelty- versus reward-factor is usually hardcoded or requires tedious tuning/annealing. In this paper, we introduce a novel measure of population-wide diversity, leveraging ideas from Determinantal Point Processes. We combine this in a principled fashion with the reward function to adapt to the degree of diversity during training, borrowing ideas from online learning. Combined with task-agnostic behavioral embeddings, we show this approach outperforms previous methods for multi-objective optimization, as well as vanilla algorithms solely optimizing for rewards.