1 paper · 1 filter
Pierre Germain, Pierre Monmarché
For optimizing a non-convex function in finite dimension, a method is to add Brownian noise to a gradient descent, allowing for transitions between basins of attractions of differe…