Timezone: »

 
Spotlight
Spectral Normalisation for Deep Reinforcement Learning: An Optimisation Perspective
Florin Gogianu · Tudor Berariu · Mihaela Rosca · Claudia Clopath · Lucian Busoniu · Razvan Pascanu

Wed Jul 21 05:40 AM -- 05:45 AM (PDT) @ None

Most of the recent deep reinforcement learning advances take an RL-centric perspective and focus on refinements of the training objective. We diverge from this view and show we can recover the performance of these developments not by changing the objective, but by regularising the value-function estimator. Constraining the Lipschitz constant of a single layer using spectral normalisation is sufficient to elevate the performance of a Categorical-DQN agent to that of a more elaborated agent on the challenging Atari domain. We conduct ablation studies to disentangle the various effects normalisation has on the learning dynamics and show that is sufficient to modulate the parameter updates to recover most of the performance of spectral normalisation. These findings hint towards the need to also focus on the neural component and its learning dynamics to tackle the peculiarities of Deep Reinforcement Learning.

Author Information

Florin Gogianu (Technical University of Cluj-Napoca)
Tudor Berariu (Imperial College London)
Mihaela Rosca (DeepMind, UCL)
Claudia Clopath (Imperial College London)
Lucian Busoniu (Technical University of Cluj-Napoca)
Razvan Pascanu (DeepMind)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors