Search papers, labs, and topics across Lattice.
1
0
3
Deep LDA implicitly conserves a (2/L) quasi-norm in deep linear networks, revealing a novel form of implicit regularization in discriminative metric learning.
