AI Paper Atlas

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

Spectral Normalisation for Deep Reinforcement Learning: An Optimisation Perspective

Florin Gogianu, Tudor Berariu, Mihaela C Rosca, Claudia Clopath, Lucian Busoniu, Razvan Pascanu

ICML 2021 · Reinforcement learning › Policy gradient and actor-critic · approach: Regularized scalable deep value learning · 12 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
Apply spectral normalisation to the penultimate layerC51 · Atari 54 games, random startsMean Human Normalised Score: 778.68 -> 1073.18Improved
Apply spectral normalisation to the penultimate layerC51 · Atari 54 games, random startsMedian Human Normalised Score: 182.26 -> 248.45Improved
Apply spectral normalisation to the penultimate layerDQN-Adam · Atari 54 games, random startsMean Human Normalised Score: 358.45 -> 719.95Improved
Apply spectral normalisation to the penultimate layerDQN-Adam · Atari 54 games, random startsMedian Human Normalised Score: 119.45 -> 178.18Improved
Apply spectral normalisation to the penultimate layerDQN-RMSprop · Atari 54 games, random startsMean Human Normalised Score: 357.36 -> 375.37Improved
Apply spectral normalisation to the penultimate layerDQN-RMSprop · Atari 54 games, random startsMedian Human Normalised Score: 102.94 -> 105.19Improved
Apply spectral normalisation to one layer rather than noneDQN across MinAtar games and optimiser settings · MinAtarMinAtar Normalised ScoreImproved
Normalise more than half the layers of a depth-5 network (ablation)DQN · MinAtarMinAtar Normalised ScoreHurt
Apply spectral normalisation to input layers or too many layers (ablation)DQN · MinAtarMinAtar Normalised ScoreHurt
Replace spectral normalisation with gradient penalty (ablation)DQN · MinAtarMinAtar Normalised ScoreHurt
Replace spectral normalisation with batch normalisation (ablation)DQN · MinAtarMinAtar Normalised ScoreHurt
Replace spectral normalisation with divOut output scaling (ablation)DQN · MinAtarMinAtar Normalised ScoreNo clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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