SUNRISE: A Simple Unified Framework for Ensemble Learning in Deep Reinforcement Learning

Abstract

Off-policy deep reinforcement learning (RL) has been successful in a range of challenging domains. However, standard off-policy RL algorithms can suffer from several issues, such as instability in Q-learning and balancing exploration and exploitation. To mitigate these issues, we present SUNRISE, a simple unified ensemble method, which is compatible with various off-policy RL algorithms. SUNRISE integrates two key ingredients: (a) ensemble-based weighted Bellman backups, which re-weight target Q-values based on uncertainty estimates from a Q-ensemble, and (b) an inference method that selects actions using the highest upper-confidence bounds for efficient exploration. By enforcing the diversity between agents using Bootstrap with random initialization, we show that these different ideas are largely orthogonal and can be fruitfully integrated, together further improving the performance of existing off-policy RL algorithms, such as Soft Actor-Critic and Rainbow DQN, for both continuous and discrete control tasks on both low-dimensional and high-dimensional environments. Our training code is available at https://github.com/pokaxpoka/sunrise.

ICML 2021 camera ready

References in corpus (10)

• DeepMind Control Suite
• Rainbow: Combining Improvements in Deep Reinforcement Learning
• Stabilizing Off-Policy Q-Learning via Bootstrapping Error Reduction
• Benchmarking Model-Based Reinforcement Learning
• Averaged-DQN: Variance Reduction and Stabilization for Deep Reinforcement Learning
• Image Augmentation Is All You Need: Regularizing Deep Reinforcement Learning from Pixels
• When to use parametric models in reinforcement learning?
• UCB Exploration via Q-Ensembles
• Exploring Model-based Planning with Policy Networks
• On the model-based stochastic value gradient for continuous reinforcement learning

Cited by in corpus (13)

• Randomized Ensembled Double Q-Learning: Learning Fast Without a Model
• On the model-based stochastic value gradient for continuous reinforcement learning
• Dreaming: Model-based Reinforcement Learning by Latent Imagination without Reconstruction
• Principled Exploration via Optimistic Bootstrapping and Backward Induction
• Improving Computational Efficiency in Visual Reinforcement Learning via Stored Embeddings
• Towards Automatic Actor-Critic Solutions to Continuous Control
• Continuous Transition: Improving Sample Efficiency for Continuous Control Problems via MixUp
• Sample Efficient Reinforcement Learning via Model-Ensemble Exploration and Exploitation
• Temporal-Difference Value Estimation via Uncertainty-Guided Soft Updates
• Ensemble Bootstrapping for Q-Learning
• Learning to Plan Optimistically: Uncertainty-Guided Deep Exploration via Latent Model Ensembles
• Measuring Progress in Deep Reinforcement Learning Sample Efficiency
• Maximum Entropy Reinforcement Learning with Mixture Policies