Stochastic Convex Function Chasing
We study the online convex optimization problem with switching costs in a stochastic environment and contrast it with the existing adversarial literature. In particular, we perform the first analysis of Convex Function Chasing (CFC) in the stochastic environment of martingales with possible drist while giving worst-case guarantees simultaneously.
- Bhuyan, N., Mukherjee, D., Wierman, A., “Best of Both Worlds: Stochastic and Adversarial Convex Function Chasing”, submitted to ACM Sigmetrics / IFIP Performance 2024, doi: arXiv:2311.00181
Adversarial Convex Function Chasing with Untrusted Advice
We are studying the adversarial online optimization problem with switching costs where black-box advice is available but cannot be completely trusted. In this area, we are trying to come up with algorithms that will break the exponential trade-off barrier between trusting the advice and safeguarding against worst-case scenarios.