Develops a new method for optimizing policies in hierarchical models.
problem Optimizing complex policies in hierarchical models.
method Applies second-order methods in the space of state-action paths.
result The natural path gradient method can be computed exactly and reflects state-space hierarchy.
HiPPO adapts skills and higher-level policies together for better transfer in hierarchical RL.
problem Sub-optimality in skill transfer when lower-level skills are fixed.
method HiPPO: a novel hierarchical policy gradient method that trains all levels of the hierarchy jointly.
result Improved robustness of skills to environment changes through training time-abstractions.
Real-world tasks are often highly structured. Hierarchical reinforcement learning (HRL) has attracted research interest as an approach for leveraging the hierarchical structure of a given task in reinforcement learning (RL). However, identifying the hierarchical policy structure that enhances the performance of RL is n…
In this work, we provide theoretical guarantees for reward decomposition in deterministic MDPs. Reward decomposition is a special case of Hierarchical Reinforcement Learning, that allows one to learn many policies in parallel and combine them into a composite solution. Our approach builds on mapping this problem into a…
Learning an optimal policy from a multi-modal reward function is a challenging problem in reinforcement learning (RL). Hierarchical RL (HRL) tackles this problem by learning a hierarchical policy, where multiple option policies are in charge of different strategies corresponding to modes of a reward function and a gati…
RHPO improves data-efficiency for hierarchical reinforcement learning.
problem High data requirements for general reinforcement learning algorithms in robotics.
method RHPO employs compositional inductive biases and task sharing mechanisms.
result RHPO enables stable and fast learning for complex domains with positive transfer.
Framework improves resilience in operations through joint long-term and short-term decision-making.
problem Resilient operations in global markets require adaptive decision rules.
method Developed a two-timescale hierarchical reinforcement learning framework.
result Framework increases mean profit by 9.2% under joint demand-supply shocks and 11.8% under prolonged shocks.
Proposes a model for multi-agent reinforcement learning with hierarchical graph attention network.
problem Limited transferability of trained policies to new multi-agent tasks.
method Uses hierarchical graph attention network for representation learning and multi-agent actor-critic for policy learning.
result Demonstrates superior performance in mixed cooperative and competitive tasks compared to existing methods.
Policy-gradient method controls multiple non-cohesive targets.
problem Controlling multiple non-cohesive targets in a decentralized manner.
method Proximal Policy Optimization for target selection and driving.
result Effective control of non-cohesive targets without prior dynamics knowledge.
MGHRL learns to generate high-level meta strategies for new tasks.
problem Efficiency and generalization in meta-RL for wide task distributions.
method Generates high-level meta strategies over subgoals, leaving subtask learning independent.
result More efficient and generalized meta-learning from past experience.
Graph Pointer Networks and hierarchical reinforcement learning solve combinatorial optimization problems like TSP.
problem Traveling Salesman Problem (TSP) with constraints.
method Graph Pointer Networks (GPNs) and hierarchical reinforcement learning.
result GPNs and hierarchical RL find optimal solutions for TSP and TSP with time windows.
Variational inference improves hierarchical imitation learning of control programs.
problem Learning structured control policies from demonstrations.
method Variational inference for discovering hierarchical structure in observation-action traces.
result Variational inference leads to more efficient and generalized control policies.
Paper tackles RL for power grid topology optimization.
problem Managing large action spaces in growing power networks.
method Hierarchical multi-agent reinforcement learning (MARL) framework.
result MARL framework outperforms single-agent RL methods.
FEN model learns fairness and efficiency in multi-agent systems.
problem Fairness in multi-agent systems for stability and productivity.
method Hierarchical reinforcement learning model with fair-efficient reward and sub-policies.
result FEN model easily learns both fairness and efficiency in multi-agent scenarios.
Unified pair trading approach using hierarchical reinforcement learning.
problem Decoupling pair selection and trading leads to limited performance.
method Hierarchical reinforcement learning framework for joint pair selection and trading.
result Unified approach outperforms existing methods on real-world stock data.
We present Multitask Soft Option Learning(MSOL), a hierarchical multitask framework based on Planning as Inference. MSOL extends the concept of options, using separate variational posteriors for each task, regularized by a shared prior. This ''soft'' version of options avoids several instabilities during training in a …
The use of imitation learning to learn a single policy for a complex task that has multiple modes or hierarchical structure can be challenging. In fact, previous work has shown that when the modes are known, learning separate policies for each mode or sub-task can greatly improve the performance of imitation learning. …
A new reinforcement learning approach using competitive primitives that specialize and specialize based on information needs.
problem Complex environments require efficient and specialized decision-making.
method Decomposes policy into competitive primitives that decide based on information needs, regularized to use minimal information.
result Improves generalization over flat and hierarchical policies.
Hierarchical decoupling improves sample efficiency for complex robots.
problem Learning long-range behaviors on complex robots.
method Two-part policy: low-level imitation and high-level transfer, with KL regularization.
result Hierarchical transfer significantly improves zero-shot high-level transfer and stabilizes learning.
Coagent policy gradient algorithms (CPGAs) are reinforcement learning algorithms for training a class of stochastic neural networks called coagent networks. In this work, we prove that CPGAs converge to locally optimal policies. Additionally, we extend prior theory to encompass asynchronous and recurrent coagent networ…
Improved cooperation between levels boosts reinforcement learning performance.
problem Training multi-level policies in hierarchical reinforcement learning.
method Modeling policy optimization as a multi-agent process and inducing cooperation between sub-policies.
result Inducing cooperation between sub-policies leads to stronger and more sample-efficient policies.
New concept of Blackwell regret for reinforcement learning with sparse rewards.
problem Sparse rewards in long horizon MDPs.
method Formalization of myopic discount factors, value functions, and policies in terms of Blackwell optimality; introduction of Blackwell regret.
result Selecting a discount factor for zero Blackwell regret becomes arbitrarily hard in long horizon MDPs.
Option Encoder compresses reinforcement learning options into a policy basis.
problem Redundant options in reinforcement learning frameworks.
method Auto-encoder framework with constrained weights to discover a policy basis.
result Option Encoder reduces the number of options while maintaining performance.
Building agents that can explore their environments intelligently is a challenging open problem. In this paper, we make a step towards understanding how a hierarchical design of the agent's policy can affect its exploration capabilities. First, we design EscapeRoom environments, where the agent must figure out how to n…
In hierarchical reinforcement learning a major challenge is determining appropriate low-level policies. We propose an unsupervised learning scheme, based on asymmetric self-play from Sukhbaatar et al. (2018), that automatically learns a good representation of sub-goals in the environment and a low-level policy that can…
The paper introduces an adjacency constraint to improve goal-conditioned HRL.
problem Training inefficiency in goal-conditioned HRL due to large action space.
method Restricting the high-level action space to a k-step adjacent region of the current state.
result The adjacency constraint preserves optimal hierarchical policies and improves HRL performance.
HRL improves open-domain dialog models by optimizing long-term conversational goals.
problem Challenges in open-domain dialog generation, including repetitive outputs, difficulty tracking conversational goals, and inappropriate text.
method Proposes VHRL, a hierarchical reinforcement learning approach using policy gradients to tune utterance-level embeddings of a variational sequence model.
result Significant improvements in human evaluation and automatic metrics over state-of-the-art dialog models.
BCPO optimizes offline RL policies by converting uncertainty into conservative bounds.
problem Offline RL's fragility under distribution shifts and model errors.
method Bayesian approach with credible lower bounds and KL regularization.
result BCPO yields an uncertainty-calibrated policy that avoids exploiting model errors.
We explore Deep Reinforcement Learning in a parameterized action space. Specifically, we investigate how to achieve sample-efficient end-to-end training in these tasks. We propose a new compact architecture for the tasks where the parameter policy is conditioned on the output of the discrete action policy. We also prop…
We address the problem of learning hierarchical deep neural network policies for reinforcement learning. In contrast to methods that explicitly restrict or cripple lower layers of a hierarchy to force them to use higher-level modulating signals, each layer in our framework is trained to directly solve the task, but acq…
This paper uses NLDT to find interpretable control rules from complex DRL policies.
problem Complex, non-interpretable policies from black-box AI methods.
method Evolutionary optimization of NLDT for hierarchical control rules.
result Interpretable control rules with similar performance to black-box DRL.
Proves EM algorithm guarantees for hierarchical imitation learning.
problem Learning hierarchical policies from expert demonstrations.
method Converts hierarchical imitation learning to latent variable model inference, analyzes EM algorithm.
result Proves EM algorithm converges to true parameter under regular conditions.
A new benchmark task for evaluating policy learning in complex, high-dimensional action spaces.
problem Lack of a commonly accepted benchmark for evaluating policy learning in hierarchical tasks with high-dimensional action spaces.
method Proposed DinerDash Gym benchmark and Decomposed Policy Graph Modelling (DPGM) algorithm.
result DPGM achieves significant improvement over baselines and effectively injects domain knowledge.
A new algorithm learns policies from batch data in hierarchical RL.
problem Learning policies from fixed batches of data without full exploration.
method Modeling RL as a two-player game with a leader-follower structure, proposing StackelbergLearner.
result StackelbergLearner achieves competitive performance in batch RL and real-world datasets.
Global convergence proved for multi-agent LQRs with hierarchical actor-critic.
problem Challenges in understanding multi-agent reinforcement learning algorithms.
method Developed a hierarchical actor-critic algorithm for partially exchangeable agents.
result Global linear convergence to optimal policy proved.
Policy-GNN optimizes GNN aggregation for diverse node iterations.
problem Optimizing GNN performance by varying aggregation iterations for different nodes.
method Policy-GNN uses a meta-policy framework with deep reinforcement learning to adaptively determine the number of aggregations for each node.
result Policy-GNN significantly outperforms state-of-the-art alternatives on real-world datasets.
In this semi-tutorial paper, we first review the information-theoretic approach to account for the computational costs incurred during the search for optimal actions in a sequential decision-making problem. The traditional (MDP) framework ignores computational limitations while searching for optimal policies, essential…
SDQL uses modular deep Q networks to efficiently learn multi-stage optimal control tasks.
problem Training complex deep reinforcement learning models for multi-stage control tasks is inefficient and unstable.
method Stacked Deep Q Learning (SDQL) with modular Q networks and backward training.
result SDQL efficiently learns optimal control policies for multi-stage tasks with high-dimensional state and action spaces.
In this work, we take a representation learning perspective on hierarchical reinforcement learning, where the problem of learning lower layers in a hierarchy is transformed into the problem of learning trajectory-level generative models. We show that we can learn continuous latent representations of trajectories, which…
Study uses machine learning to analyze state drug policies and reduce overdose deaths.
problem Epidemic opioid overdose rates and ineffective state-level policies.
method Hierarchical clustering of 138 binomial variables to generate policy bundles, then regression analysis.
result Balancing certain policies leads to reduced overdose deaths, but only after second year.
A new method solves bilevel optimization problems in competitive Markov games.
problem Capturing competitive structures in RL with multiple interacting policies.
method Penalty-augmented Nikaido-Isoda descent-ascent (PANDA) method.
result PANDA converges to stationary points without convexity assumptions.
Develops Heuristic Portfolio Optimization (HPO) as an information-restricted projection of Markowitz/tangency solution
problem Practitioners allocate capital with forecast-light rules like equal weight, inverse volatility, risk parity, HRP, and RA-HRP
method Implies-return principle and fixed-tree cluster-Sharpe recursion
result Formalizes HPO maps, proves defect equals squared inefficiency, and identifies nodewise alphas as policy-gradient coordinates
DSE learns transferable skills across changing dynamics and goals.
problem Learning transferable skills across different reinforcement learning tasks.
method Variational inference for multi-task reinforcement learning with shared and task-specific latent spaces.
result Policies can generalize to unseen dynamics and goals conditions.
MAVEN improves multi-agent exploration by hybridizing value and policy-based methods.
problem Exploration and suboptimality in complex multi-agent environments.
method MAVEN combines value and policy-based approaches with a latent space for hierarchical control.
result MAVEN achieves significant performance improvements on challenging multi-agent tasks.
New hierarchical RL agent learns to generalize in complex multi-agent games.
problem Current RL methods struggle to generalize to unseen opponents in multi-agent games.
method Proposed a hierarchical RL architecture grounded in game-theoretic structure.
result Hierarchical agent generalizes to unseen opponents, while baselines fail.
H-ReIL learns to drive safely in near-accident scenarios.
problem Driving safely in high-risk near-accident situations.
method Hierarchical RL and IL approach.
result High-level policy switches between low-level policies for safe driving.
We introduce Compositional Imitation Learning and Execution (CompILE): a framework for learning reusable, variable-length segments of hierarchically-structured behavior from demonstration data. CompILE uses a novel unsupervised, fully-differentiable sequence segmentation module to learn latent encodings of sequential d…
Method solves long-horizon robotic tasks via imitation and reinforcement learning.
problem Long-horizon robotic tasks with complex sequences of actions.
method Two-phase approach: imitation learning followed by reinforcement learning.
result Method can scale to challenging long-horizon tasks.