Course webpage: www.cs.umd.edu/...
0:15 Agenda 1:17 ERM (non-convex opt, key practical observations, over-parameterized regime) 9:30 Q&A 13:03 Examples 16:30 warm-up 25:34 Under-parameterized regime (essential non-convexity) 34:55 Q&A 36:27 PL Conditions 41:00 Tangent kernel 46:11 Proof 50:58 Q&A 52:36 Informal convergence result 56:00 Example 1:00:30 Intuition (over parameterized system has good condition number) 1:02:10 Q&A 1:03:56 Convergence Proof 1:10:14 Q&A
thanks and very benefit for me
Thanks for the excellent lecture! Is there a site for the scribe notes ?
explanation of: "Loss landscapes and optimization in over-parameterized non-linear systems and neural networks"
24:18 The loss function is sometime defined by an L and sometime edfines by the caligraohic L, are they the same? thank you very much !
Does anyone have the reference for the matrix multiplication at 1:03:10?
👏👏👏
Пікірлер: 7
0:15 Agenda 1:17 ERM (non-convex opt, key practical observations, over-parameterized regime) 9:30 Q&A 13:03 Examples 16:30 warm-up 25:34 Under-parameterized regime (essential non-convexity) 34:55 Q&A 36:27 PL Conditions 41:00 Tangent kernel 46:11 Proof 50:58 Q&A 52:36 Informal convergence result 56:00 Example 1:00:30 Intuition (over parameterized system has good condition number) 1:02:10 Q&A 1:03:56 Convergence Proof 1:10:14 Q&A
thanks and very benefit for me
Thanks for the excellent lecture! Is there a site for the scribe notes ?
explanation of: "Loss landscapes and optimization in over-parameterized non-linear systems and neural networks"
24:18 The loss function is sometime defined by an L and sometime edfines by the caligraohic L, are they the same? thank you very much !
Does anyone have the reference for the matrix multiplication at 1:03:10?
👏👏👏