There was quickly rising curiosity in creating strategies for meta-learning inside deep RL. Though there was substantive progress towards such ‘meta-reinforcement studying,’ analysis on this space has been held again by a scarcity of benchmark duties. Within the current work, we intention to ease this downside by introducing (and open-sourcing) Alchemy, a helpful new benchmark surroundings for meta-RL, together with a set of study instruments.
Recreation concept as an engine for large-scale information evaluation
