Phase transition LoFi

Open gerdm opened this issue 2 years ago • 1 comments

Plot performance vs L for L=1:20 for LOFI on FMNIST and UCI to see if there is a phase transition (cf Larsen, and Li)

Mar 27 '23 13:03 gerdm

B. W. Larsen, S. Fort, N. Becker, and S. Ganguli, “How many degrees of freedom do we need to train deep networks: a loss landscape perspective,” in ICLR, 2022 [Online]. Available: http://arxiv.org/abs/2107.05802

C. Li, H. Farkhoor, R. Liu, and J. Yosinski, “Measuring the Intrinsic Dimension of Objective Landscapes,” in ICLR, 2018 [Online]. Available: http://arxiv.org/abs/1804.08838

Mar 27 '23 21:03 murphyk