Reproduce “Discriminative k-shot learning using probabilistic models” 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
Matthias Bauer, Mateo Rojas-Carulla, Jakub Bartłomiej Świątkowski, Bernhard Schölkopf, Richard E. Turner
The method adapts classifiers to new classes from few labeled examples by freezing old-class image features and using a Gaussian prior fitted to old classifier weights. It assumes old and new class weights share a distribution. On 5-way, 5-shot miniImageNet, ResNet-34 features reach 73.9 ± 0.3% accuracy, versus 68.2 ± 0.7% for Prototypical Networks; the comparison does not isolate prior contribution.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.