LCA: Loss Change Allocation for Neural Network Training
Janice Lan, Rosanne Liu, Hattie Zhou, Jason Yosinski
–Neural Information Processing Systems
This rich view shows which parameters are responsible for decreasing or increasing the loss during training, orwhich parameters "help" or"hurt" the network'slearning, respectively.
Neural Information Processing Systems
Feb-14-2026, 10:55:56 GMT