Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
1
0
0
1
Wider networks enable greedy layer-wise training to match or even surpass end-to-end backpropagation in self-supervised learning, driven by superior representational geometry.