When searching for policies, reward-sparse environments often lack sufficient information about which behaviors to improve upon or avoid. In such environments, the policy search process is bound to blindly search for reward-yielding transitions and no early reward can bias this search in one direction or another. A way to overcome this is to use intrinsic motivation in order to explore new transitions until a reward is found. In this work, we use a recently proposed definition of intrinsic motivation, Curiosity, in an evolutionary policy search method. We propose Curiosity-ES, an evolutionary strategy adapted to use Curiosity as a fitness metric. We compare Curiosity with Novelty, a commonly used diversity metric, and find that Curiosity can generate higher diversity over full episodes without the need for an explicit diversity criterion and lead to multiple policies which find reward.
Although transfer learning is considered to be a milestone in deep reinforcement learning, the mechanisms behind it are still poorly understood. In particular, predicting if knowledge can be transferred between two given tasks is still an unresolved problem. In this work, we explore the use of network distillation as a feature extraction method to better understand the context in which transfer can occur. Notably, we show that distillation does not prevent knowledge transfer, including when transferring from multiple tasks to a new one, and we compare these results with transfer without prior distillation. We focus our work on the Atari benchmark due to the variability between different games, but also to their similarities in terms of visual features.
Learning a good state representation is a critical skill when dealing with multiple tasks in Reinforcement Learning as it allows for transfer and better generalization between tasks. However, defining what constitute a useful representation is far from simple and there is so far no standard method to find such an encoding. In this paper, we argue that distillation -- a process that aims at imitating a set of given policies with a single neural network -- can be used to learn a state representation displaying favorable characteristics. In this regard, we define three criteria that measure desirable features of a state encoding: the ability to select important variables in the input space, the ability to efficiently separate states according to their corresponding optimal action, and the robustness of the state encoding on new tasks. We first evaluate these criteria and verify the contribution of distillation on state representation on a toy environment based on the standard inverted pendulum problem, before extending our analysis on more complex visual tasks from the Atari and Procgen benchmarks.
Deep reinforcement learning policies, despite their outstanding efficiency in simulated visual control tasks, have shown disappointing ability to generalize across disturbances in the input training images. Changes in image statistics or distracting background elements are pitfalls that prevent generalization and real-world applicability of such control policies. We elaborate on the intuition that a good visual policy should be able to identify which pixels are important for its decision, and preserve this identification of important sources of information across images. This implies that training of a policy with small generalization gap should focus on such important pixels and ignore the others. This leads to the introduction of saliency-guided Q-networks (SGQN), a generic method for visual reinforcement learning, that is compatible with any value function learning method. SGQN vastly improves the generalization capability of Soft Actor-Critic agents and outperforms existing stateof-the-art methods on the Deepmind Control Generalization benchmark, setting a new reference in terms of training efficiency, generalization gap, and policy interpretability.
Over the past few years, the acceleration of computing resources and research in deep learning has led to significant practical successes in a range of tasks, including in particular in computer vision. Building on these advances, reinforcement learning has also seen a leap forward with the emergence of agents capable of making decisions directly from visual observations. Despite these successes, the over-parametrization of neural architectures leads to memorization of the data used during training and thus to a lack of generalization. Reinforcement learning agents based on visual inputs also suffer from this phenomenon by erroneously correlating rewards with unrelated visual features such as background elements. To alleviate this problem, we introduce a new regularization technique consisting of channel-consistent local permutations (CLOP) of the feature maps. The proposed permutations induce robustness to spatial correlations and help prevent overfitting behaviors in RL. We demonstrate, on the OpenAI Procgen Benchmark, that RL agents trained with the CLOP method exhibit robustness to visual changes and better generalization properties than agents trained using other state-of-the-art regularization techniques. We also demonstrate the effectiveness of CLOP as a general regularization technique in supervised learning.
Neurogenesis in ANNs is an understudied and difficult problem, even compared to other forms of structural learning like pruning. By decomposing it into triggers and initializations, we introduce a framework for studying the various facets of neurogenesis: when, where, and how to add neurons during the learning process. We present the Neural Orthogonality (NORTH*) suite of neurogenesis strategies, combining layer-wise triggers and initializations based on the orthogonality of activations or weights to dynamically grow performant networks that converge to an efficient size. We evaluate our contributions against other recent neurogenesis works with MLPs.
Deep neural networks have demonstrated their ability to automatically extract meaningful features from data. However, in supervised learning, information specific to the dataset used for training, but irrelevant to the task at hand, may remain encoded in the extracted representations. This remaining information introduces a domain-specific bias, weakening the generalization performance. In this work, we propose splitting the information into a task-related representation and its complementary context representation. We propose an original method, combining adversarial feature predictors and cyclic reconstruction, to disentangle these two representations in the single-domain supervised case. We then adapt this method to the unsupervised domain adaptation problem, consisting of training a model capable of performing on both a source and a target domain. In particular, our method promotes disentanglement in the target domain, despite the absence of training labels. This enables the isolation of task-specific information from both domains and a projection into a common representation. The task-specific representation allows efficient transfer of knowledge acquired from the source domain to the target domain. In the single-domain case, we demonstrate the quality of our representations on information retrieval tasks and the generalization benefits induced by sharpened task-specific representations. We then validate the proposed method on several classical domain adaptation benchmarks and illustrate the benefits of disentanglement for domain adaptation.
Several algorithms have been proposed to sample non-uniformly the replay buffer of deep Reinforcement Learning (RL) agents to speed-up learning, but very few theoretical foundations of these sampling schemes have been provided. Among others, Prioritized Experience Replay appears as a hyperparameter sensitive heuristic, even though it can provide good performance. In this work, we cast the replay buffer sampling problem as an importance sampling one for estimating the gradient. This allows deriving the theoretically optimal sampling distribution, yielding the best theoretical convergence speed. Elaborating on the knowledge of the ideal sampling scheme, we exhibit new theoretical foundations of Prioritized Experience Replay. The optimal sampling distribution being intractable, we make several approximations providing good results in practice and introduce, among others, LaBER (Large Batch Experience Replay), an easy-to-code and efficient method for sampling the replay buffer. LaBER, which can be combined with Deep Q-Networks, distributional RL agents or actor-critic methods, yields improved performance over a diverse range of Atari games and PyBullet environments, compared to the base agent it is implemented on and to other prioritization schemes.
We consider the problem of knowledge transfer when an agent is facing a series of Reinforcement Learning (RL) tasks. We introduce a novel metric between Markov Decision Processes and establish that close MDPs have close optimal value functions. Formally, the optimal value functions are Lipschitz continuous with respect to the tasks space. These theoretical results lead us to a value transfer method for Lifelong RL, which we use to build a PAC-MDP algorithm with improved convergence rate. We illustrate the benefits of the method in Lifelong RL experiments.