3 papers
cs.LG2026
When Softmax Fails at the Top: Extreme Value Corrections for InfoNCE
Melihcan Erol, Suat Evren, Oktay Ozel +3
InfoNCE is the standard contrastive learning objective, but its softmax form is not only a computational convenience: it also encodes a statistical assumption about how the top-sco…
cs.LG2024
Operator SVD with Neural Networks via Nested Low-Rank Approximation
J. Jon Ryu, Xiangxiang Xu, H. S. Melihcan Erol +3
Computing eigenvalue decomposition (EVD) of a given linear operator, or finding its leading eigenvalues and eigenfunctions, is a fundamental task in many machine learning and scien…
math.PR2021
Variations on Hammersley's interacting particle process
Arda Atalik, H. S. Melihcan Erol, Gökhan Yıldırım +1
The longest increasing subsequence problem for permutations has been studied extensively in the last fifty years. The interpretation of the longest increasing subsequence as the lo…