Reproduce “The Stability of Online Algorithms in Performative Prediction” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Gabriele Farina, Juan Carlos Perdomo
The authors show that a uniform lottery over a deterministic no-regret online algorithm’s iterates is stable in expectation, even when deployment changes the data distribution arbitrarily. The bounded-loss guarantee applies to the mixture, not each iterate. For projected online gradient descent on convex losses with gradients bounded by G over a domain of diameter D, its expected stability gap is at most 3GD/(2√T).
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.