Titouan Vayer

Unifying optimal transport with generative modeling techniques reveals deeper insights into their mathematical underpinnings and practical applications.

Rescaling ReLU networks using a geometrically-motivated path-conditioning strategy can significantly speed up training by aligning a kernel in the path-lifting space.

Gromov-Wasserstein optimal transport plans can be sparse and permutation-supported under certain conditions, challenging the intuition that they are always dense.