Supplementary Information: Meta-ReinforcementLearningwith Self-ModifyingNetworks 9 Optimization

Neural Information Processing Systems 

General information: As specified in section3 and 5, we test a single model definition for all experiments in this work, with one layer of dynamic weightsWt. This layer consists in a dense matrix ofsizen Nwith learnt orrandom initialization.

Similar Docs  Excel Report  more

TitleSimilaritySource
None found