@Allen_Schmaltz But via self-labeling, which sounds like classic semi-supervised learning. If you are willing to do some form of active learning, then this is similar to Andreas Krause's safe RL work (e.g., https://t.co/Lt6Cz1kOBN).
AI Weekly's analysis
→
- The paper extends Lyapunov stability verification and pairs it with statistical dynamics models to produce control policies with provable stability certificates.
- Under a Gaussian process prior, the authors prove that data collection itself can be done safely while expanding the certified safe region of the state space.
- In simulation, the method optimizes a neural network policy on an inverted pendulum without the pendulum ever falling down.
Read full analysis →