Search papers, labs, and topics across Lattice.
1
0
2
Rescaling ReLU networks using a geometrically-motivated path-conditioning strategy can significantly speed up training by aligning a kernel in the path-lifting space.
