Single Level Optimization

Single-level optimization focuses on finding the optimal solution to a single objective function, a fundamental problem across numerous scientific fields. Current research emphasizes developing efficient algorithms, such as those based on stochastic gradient descent and adaptive step-size methods, to improve convergence speed and stability, particularly for large-scale problems. This area is crucial for advancing machine learning, where single-level optimization underpins many model training processes, and also finds applications in diverse fields like operations research and engineering design. Recent work also explores the implicit biases of these algorithms and methods for discovering multiple optima.