Research map / Reinforcement learning / Visual and representation RL
Deep Laplacian-based Options for Temporally-Extended Exploration
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Fully online option discovery (no separate pretraining phase) with generalized Laplacian options (ablation) | DCEO on pixel-based Nine rooms, Maze, Rubik's cube · Pixel-based reward maximization | reward performance vs baselines | No clear effect |
| Using options (DCEO) for exploration vs count-based, RND, DIAYN, epsilon-z-greedy baselines | DCEO in Maze reward maximization · Maze pixel-based | reward performance gap | Helped |
| Option-based exploration in non-stationary Nine rooms (changed goal and topology) | DCEO · Nine rooms non-stationary | adaptation speed after transfer | Helped |
| Removing the Laplacian options from exploration in DIAYN-style mutual-information options | DIAYN baseline · Nine rooms reward maximization | reward vs DDQN | Hurt |
| Using the generalized Laplacian objective instead of the original graph-drawing Laplacian objective (ablation) | Eigenfunction approximation · Comparison of Laplacian objectives | accuracy of eigenfunction approximation | Helped |
| Sharing weights between option/Laplacian networks and the main reward-maximizing network removed (separate networks) | DCEO · Design choice | reward performance | No clear effect |
| Random-restart removal: DCEO learns eigenfunctions without random restarts on pixel inputs | DCEO state coverage · Pixel-based state coverage | number of states visited | Helped |
| Using DCEO options on Montezuma's Revenge and MiniWorld 3D navigation with DDQN/Rainbow backbones | DCEO · MiniWorld navigation and Montezuma's Revenge | reward | Helped |
Closest papers by meaning
- Addressing Sparse-Rewards in RL with Scalable Hierarchical Novel Eigen Options (NeurIPS 2026 · similarity 0.92)
- Online Laplacian-Based Representation Learning in Reinforcement Learning (ICML 2025 · similarity 0.91)
- Proper Laplacian Representation Learning (ICLR 2024 · similarity 0.90)
- The Laplacian Keyboard: Beyond the Linear Span (NeurIPS 2026 · similarity 0.90)
- Stochastic Q-learning for Large Discrete Action Spaces (ICML 2024 · similarity 0.87)
- Enhancing Value Function Estimation through First-Order State-Action Dynamics in Offline Reinforcement Learning (ICML 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.