AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

Learning Generalizable Agents via Saliency-guided Features Decorrelation

Sili Huang, Yanchao Sun, Jifeng Hu, Siyuan Guo, Hechang Chen, Yi Chang, Lichao Sun, Bo Yang

NeurIPS 2023 · Reinforcement learning › Visual and representation RL · approach: Robust visual encoders for RL · 3 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (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 changedWhereMeasuredEffect
SGFD vs strongest baseline TED/AMBS under background noise (total over 5 tasks)SGFD · DMC background noisetotal reward: 3734.9 -> 3929.5Improved
SGFD vs AMBS under interpolation of robot parametersSGFD · DMC interpolationtotal reward: 1975.6 -> 2273.4Improved
SGFD vs AMBS under extrapolation of robot parametersSGFD · DMC extrapolationtotal reward: 1857.6 -> 2253.1Improved
SGFD vs AMBS on walker-walk with background noiseSGFD · DMC walker-walk backgroundreward: 926.7 -> 959.1Improved
SGFD vs DBC on MT-finger-spin interpolationSGFD · MT-finger-spinreward: 221.6 -> 234.1Improved
remove RFF from decorrelation (ablation)SGFD · Cheetah-run background and MT-cheetah-runtest rewardHurt
decorrelation without RFF vs no decorrelation (ablation)SGFD ablation · Cheetah-run and MT-cheetah-runtest rewardImproved
remove saliency-guided model (global decorrelation of all features) (ablation)SGFD · Cheetah-run and MT-cheetah-runtest rewardHurt
saliency-guided weighting vs weightingSGFD · Picking task Pearson correlation of mass feature s10correlationImproved
reweight samples with SGFDSGFD · Visualization of 'fall' states in two environmentsshare of similar states: 50% and 35.4% -> approximately 42%Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy