Reproduce “MI-Hyper: Generating Multiple Instance Models for Weight Space Transfer Learning in Histopathology via Mixed Tasks Pretrained HyperNetwork” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Xitong Ling, Yousef Kotp, Weiming Chen, Yining Li, Jiawen Li, Minxi Ouyang, Xinyi Guo, Mingxi Fu and 4 more
MI-Hyper builds a whole-slide pathology predictor from task and class descriptions, generating attention queries and classifier weights that select and combine patch evidence. Across 308 tasks from 52 datasets, it achieved mean macro one-vs-rest AUC of 58.24%, versus 56.38% for MI-Zero. This average concerns held-out task IDs, not an entirely unseen patient population.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.