SUSD: Structured Unsupervised Skill Discovery through State Factorization

1. 1Apply Mutual Information to skill latent variables and states
2. 2Configure State Factorization to capture dynamic behaviors
3. 3Build a Distance-Maximizing Skill Discovery model
4. 4Test the performance of the model on various tasks
5. 5Run experiments to evaluate the discovery of task-relevant skills

AI engineers and researchers on a team can benefit from this approach to improve the autonomy of their agents, while data scientists can apply this method to discover new insights in complex systems

💡 State Factorization can help capture dynamic task-relevant behaviors beyond simple static skills

💡 Discover dynamic skills without rewards using Structured Unsupervised Skill Discovery!