2021 (modified: 26 Feb 2022)ICML 2021Readers: Everyone
Abstract:Recent results in supervised learning suggest that while overparameterized models have the capacity to overfit, they in fact generalize quite well. We ask whether the same phenomenon occurs for off...