🔹 Theoretically ECP is a no-regret algorithm.
🔹 Outperforms state-of-the-art algorithms, including Lipschitz, Bayesian, bandits, and evolutionary approaches.
📜 Paper: arxiv.org/abs/2502.042...
💻 Code: github.com/fouratifares... – try it out, and share your thoughts!
🔹 ECP is optimized for expensive functions—excels with low function call budgets.
🔹 ECP is empirically tested across 30 multi-dimensional non-convex synthetic and real-world optimization problems.
🚀 Tackling non-convex optimization? Stuck in local optima? Need a global optimizer for black-box functions?
Introducing ECP—a powerful global optimization algorithm (AISTATS ’25)! 🔥
🔹 ECP focuses on regions with the highest potential for optimality.
PhD candidate in RL. Can I join?
