Logit-Based Losses Limit the Effectiveness of Feature Knowledge Distillation
Paper • 2511.14981 • Published
Nick Cooper
NickCombi
·
AI & ML interests
I enjoy using discrete mathematics to study deep neural networks. Generally, I'm interested in how graph theory, combinatorics, and topological data analysis can provide insights into the structure of intelligence.
Recent Activity
authored a paper 1 day ago
Logit-Based Losses Limit the Effectiveness of Feature Knowledge Distillation authored a paper 1 day ago
On the Architectural Complexity of Neural NetworksOrganizations
authored 2
papers 1 day ago