AI Paper Atlas

Research map / Reinforcement learning / Policy gradient and actor-critic

Adapting to Reward Progressivity via Spectral Reinforcement Learning

Michael Dann, John Thangarajah

ICLR 2021 · Reinforcement learning › Policy gradient and actor-critic · approach: Temporal-difference and discounting theory · 1 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
Use Spectral DQN instead of Pop-Art or DQN+TCPonglantis · Ponglantismean episodic returnImproved
Use Spectral DQN instead of Pop-Art or DQN+TCExponential Pong · Exponential Pongmean episodic returnImproved
Use Spectral DQN instead of both baselinesAtari agents · Bowlingmean episodic returnImproved
Use Spectral DQN instead of both baselinesAtari agents · Chopper Commandmean episodic returnImproved
Use Spectral DQN instead of both baselinesAtari agents · Video Pinballmean episodic returnImproved
Use Spectral DQN instead of DQN+TCAtari agents · Centipedemean episodic returnHurt
Use Spectral DQN instead of the baselinesAtari agents · the other standard Atari gamesmean episodic returnNo clear effect
Use DQN+TC instead of Pop-Artunclipped-reward DQN · Ponglantismean episodic returnImproved
Set spectral-frequency loss weights to b^i instead of balancing frequencies (ablation)Spectral DQN · Ponglantis and Exponential Pongmean episodic returnHurt
Remove target-scale correction for hidden-layer gradients (set w_i = 1) (ablation)Spectral DQN · standard Atari gamesmean episodic return: 4 of the 6 gamesHurt
Scale Atlantis-phase perceived rewards down by a factor 10DQN+TC · Ponglantis Easiermean episodic returnImproved
Scale Atlantis-phase perceived rewards down by a factor 100Pop-Art · Ponglantis Even Easiermean episodic returnImproved

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