TOYBOX creates new Atari environments for deep RL experiments.
problem Challenges in evaluating deep RL agent behavior.
method Designs and opensource releases a subset of Atari environments.
result TOYBOX enables experiments impossible in other environments.
Reinforcement learning is concerned with identifying reward-maximizing behaviour policies in environments that are initially unknown. State-of-the-art reinforcement learning approaches, such as deep Q-networks, are model-free and learn to act effectively across a wide range of environments such as Atari games, but requ…
A new reward learning module improves imitation learning in high-dimensional environments.
problem Challenges in high-dimensional environments for imitation learning.
method Generative model to generate intrinsic reward signals.
result Our method outperforms state-of-the-art IRL methods on Atari games.
CuLE accelerates reinforcement learning on GPUs.
problem Limited CPU-GPU communication bandwidth in Atari Learning Environment.
method CUDA port of ALE, leveraging GPU parallelization and direct frame rendering.
result Up to 155M frames per hour on a single GPU, accelerating reinforcement learning.
A new neural network model improves reinforcement learning efficiency.
problem Improving sample efficiency in reinforcement learning.
method Proposes a neural network architecture using VQ-VAE and convolutional LSTM for world modeling.
result Shows comparable performance to SimPLe with significantly smaller model.
This paper benchmarks batch RL algorithms on Atari, finding DQN and partially-trained policies perform best.
problem Deep RL algorithms fail in batch setting.
method Benchmarked batch RL algorithms on Atari using a single partially-trained policy.
result Many batch RL algorithms underperform DQN and partially-trained policies.
Study finds reinforcement learning performance plateaus due to environmental interference.
problem Catastrophic interference hinders sample efficiency in reinforcement learning.
method Empirical study in ALE, controlled experiments, analysis of prediction errors.
result Interference causes performance plateaus and degrades policies used to reach them.
Method learns state representations without supervision for Atari games.
problem Learning state representations without supervision.
method Maximizes mutual information across features of neural encoder.
result New benchmark for evaluating representation learning models.
Deep RL approach improves MIS for complex environments.
problem Improving off-policy evaluation for complex environments.
method Uses successor representation from deep RL to decouple reward and dynamics.
result Empirically stable and applicable to high-dimensional domains.
Paper presents a new method for Bayesian deep learning that scales to Atari games.
problem Training neural networks on complex environments like Atari games is challenging.
method Adapted temporal difference Q-learning to work with Bayesian inference.
result TAGI allows for analytical inference of neural network parameters, achieving performance comparable to gradient-based methods.
DreamerV2 learns Atari game behaviors from a world model, achieving human-level performance.
problem Learning complex behaviors in Atari games from limited data.
method DreamerV2 uses a world model with discrete representations to predict behaviors in a compact latent space.
result Achieves human-level performance on 55 Atari tasks.
New loss function equivalence reveals PER's uniform sampling can be improved.
problem Improving Prioritized Experience Replay (PER) for better learning efficiency.
method Transforming non-uniformly sampled data loss functions into uniformly sampled ones.
result Some environments can replace PER with a new loss function without performance loss.
MuZero learns models to master complex games without domain knowledge.
problem Mastering complex, real-world domains with unknown dynamics.
method Combining tree-based search with learned models.
result Achieves superhuman performance in diverse domains.
Recent reinforcement learning (RL) approaches have shown strong performance in complex domains such as Atari games, but are often highly sample inefficient. A common approach to reduce interaction time with the environment is to use reward shaping, which involves carefully designing reward functions that provide the ag…
DeepMDP simplifies complex observations into continuous latent states.
problem Learning from high-dimensional observations in reinforcement learning.
method Trains a DeepMDP model that predicts rewards and next latent states.
result Optimization of DeepMDP objectives ensures quality of latent space and environment model.
Model-free reinforcement learning (RL) can be used to learn effective policies for complex tasks, such as Atari games, even from image observations. However, this typically requires very large amounts of interaction -- substantially more, in fact, than a human would need to learn the same games. How can people learn so…
A new HRL algorithm learns and exploits multiple subgoals for faster exploration.
problem Sparse reward problem in reinforcement learning.
method Multi-goal HRL algorithm with Manager and Worker policies.
result Significantly improved exploration efficiency with reduced training time.
Efficient exploration in complex environments remains a major challenge for reinforcement learning. We propose bootstrapped DQN, a simple algorithm that explores in a computationally and statistically efficient manner through use of randomized value functions. Unlike dithering strategies such as epsilon-greedy explorat…
Models that can simulate how environments change in response to actions can be used by agents to plan and act efficiently. We improve on previous environment simulators from high-dimensional pixel observations by introducing recurrent neural networks that are able to make temporally and spatially coherent predictions f…
Reproducibility in reinforcement learning is challenging: uncontrolled stochasticity from many sources, such as the learning algorithm, the learned policy, and the environment itself have led researchers to report the performance of learned agents using aggregate metrics of performance over multiple random seeds for a …
Paper proposes a new method for better estimating continuous distributions in RL.
problem Challenges in parameterizing estimated distributions for better approximation of true continuous distribution.
method Proposes fully parameterized quantile function with fraction and value networks.
result Significantly outperforms existing distributional RL algorithms on 55 Atari Games.
Simple object representations improve model-free RL performance.
problem Current reinforcement learning agents lack object recognition.
method Used simple, feature-engineered object representations with the Rainbow model.
result Object representations significantly boost performance on Atari games.
CURL uses contrastive learning to improve reinforcement learning performance.
problem Improving reinforcement learning performance on complex tasks.
method Contrastive learning to extract high-level features from raw pixels, followed by off-policy control.
result CURL outperforms prior methods on DeepMind Control Suite and Atari Games.
A new method for prioritizing sequences in reinforcement learning improves learning efficiency.
problem Efficient learning in reinforcement learning algorithms.
method Prioritized Sequence Experience Replay (PSER) for prioritizing sequences of experience.
result PSER is guaranteed to converge faster than PER and substantially improves performance.
MB-DQN uses different backup lengths for improved reinforcement learning.
problem Improving reinforcement learning with multi-step returns.
method Integrates multi-step returns into bootstrapped DQN with different backup lengths.
result MB-DQN maintains advantages of different backup lengths.
AR-A3C improves A3C's robustness to noisy environments.
problem Neural networks are vulnerable to noise from unexpected sources in reinforcement learning.
method Introduced an adversarial agent to make the learning process more robust.
result AR-A3C outperforms A3C in both clean and noisy environments.
Self-Predictive Representations improves data-efficient reinforcement learning from limited interaction.
problem Efficient reinforcement learning from limited data.
method Train agents to predict future latent state representations using self-supervised objectives.
result Achieves a median human-normalized score of 0.415 on Atari with 100k steps of interaction, 55% improvement over previous state-of-the-art.
A method for self-supervised representation learning in partially observable environments.
problem Sparse rewards and stochasticity in partially observable environments.
method World model in latent space for estimating missing information.
result Significant improvement in exploration compared to prior work.
For sophisticated reinforcement learning (RL) systems to interact usefully with real-world environments, we need to communicate complex goals to these systems. In this work, we explore goals defined in terms of (non-expert) human preferences between pairs of trajectory segments. We show that this approach can effective…
A new method prioritizes and recycles experiences for better reinforcement learning.
problem Improving reinforcement learning efficiency by prioritizing and recycling experiences.
method Double-prioritized state-recycled (DPSR) experience replay.
result DPSR achieved state-of-the-art results in Atari games, outperforming original and prioritized methods.
BYOL-Explore learns to explore visually-rich environments by predicting world dynamics.
problem Exploration in visually complex environments.
method Optimizes a single prediction loss in latent space to learn world representation, dynamics, and exploration policy.
result Achieves superhuman performance on Atari games with simpler design.
DPFRL uses particle filters for decision making with complex visual observations.
problem Decision making with partial complex visual observations.
method Discriminative Particle Filter Reinforcement Learning (DPFRL) with a differentiable particle filter in the neural network policy.
result DPFRL outperforms state-of-the-art POMDP RL models in complex visual observation tasks.
MDP Playground tests RL agents across various dimensions for better understanding and debugging.
problem Understanding and debugging reinforcement learning agents across diverse environments and dimensions.
method Controlled testbed with adjustable dimensions for different RL challenges.
result Insights into agent performance and interaction with various dimensions.
FICM uses motion features to encourage exploration in DRL environments.
problem Exploiting novelty in environments with moving objects.
method Flow-based intrinsic curiosity module (FICM) leveraging motion features between consecutive frames.
result FICM encourages exploration in environments featuring moving objects.
Curiosity enhanced by audio-visual associations improves learning efficiency.
problem Challenges in reinforcement learning, especially predicting the future.
method Exploits multiple modalities (audio and vision) to predict novel associations.
result Improves exploration and learning efficiency in various environments.
Proposes ABIDES-Gym for financial markets simulation.
problem Training multi-agent systems in financial markets.
method Wrapping DEMAS into OpenAI Gym framework.
result Developed two financial markets OpenAI Gym environments.
A reinforcement learning model uses attention to make actions interpretable.
problem Creating interpretable reinforcement learning models.
method Soft attention mechanism to focus on task-relevant information.
result The model achieves competitive performance on Atari tasks and is interpretable.
Complex environments and tasks pose a difficult problem for holistic end-to-end learning approaches. Decomposition of an environment into interacting controllable and non-controllable objects allows supervised learning for non-controllable objects and universal value function approximator learning for controllable obje…
New algorithms improve reinforcement learning stability and performance.
problem Stability issues in TD learning algorithms with function approximation and off-policy sampling.
method Developed and adapted emphatic temporal difference (ETD(λ)) algorithms for deep reinforcement learning. result Demonstrated improved performance in Atari games and small problems.
Many practical environments contain catastrophic states that an optimal agent would visit infrequently or never. Even on toy problems, Deep Reinforcement Learning (DRL) agents tend to periodically revisit these states upon forgetting their existence under a new policy. We introduce intrinsic fear (IF), a learned reward…
Learning to control an environment without hand-crafted rewards or expert data remains challenging and is at the frontier of reinforcement learning research. We present an unsupervised learning algorithm to train agents to achieve perceptually-specified goals using only a stream of observations and actions. Our agent s…
Adapts exploration for better learning progress in distributed agents.
problem Optimizing exploration in distributed reinforcement learning environments.
method Uses a non-stationary multi-armed bandit to dynamically adapt data generation parameters.
result Unified approach produces comparable results to per-task tuning at lower cost.
Reinforcement learning algorithms rely on carefully engineering environment rewards that are extrinsic to the agent. However, annotating each environment with hand-designed, dense rewards is not scalable, motivating the need for developing reward functions that are intrinsic to the agent. Curiosity is a type of intrins…
CRL uses clustering to enhance exploration in RL, improving performance in diverse tasks.
problem Challenges in RL with large state spaces and sparse rewards.
method CRL employs clustering to guide exploration based on novelty and quality in neighboring areas.
result CRL outperforms other methods in various RL tasks.
Paper develops an online regularization framework for RL to decorrelate features.
problem Learning good representations in reinforcement learning.
method Online regularization framework using Gramian of features.
result Significant improvement in sample efficiency on Atari 2600 games.
Recent developments in deep reinforcement learning have enabled the creation of agents for solving a large variety of games given a visual input. These methods have been proven successful for 2D games, like the Atari games, or for simple tasks, like navigating in mazes. It is still an open question, how to address more…
Deep Reinforcement Learning (RL) recently emerged as one of the most competitive approaches for learning in sequential decision making problems with fully observable environments, e.g., computer Go. However, very little work has been done in deep RL to handle partially observable environments. We propose a new architec…
The paper proposes a framework to reason about object dynamics for faster reinforcement learning.
problem Current reinforcement learning approaches lack prior knowledge about the environment, limiting efficiency.
method Integrates object dynamics and behavior into reinforcement learning to improve efficiency.
result Demonstrates the need for reasoning about object behavior and dynamics, leading to faster learning.