Search papers, labs, and topics across Lattice.
Affiliation:
1
0
Gradient descent can slash held-out error by up to 44% compared to Adam in low-rank matrix recovery, revealing the critical role of optimizer choice in performance outcomes.
