Reproduce “Lifelong Machine Learning with Deep Streaming Linear Discriminant Analysis” 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
Tyler L. Hayes, Christopher Kanan
Deep SLDA adapts image classification to one-pass streams by keeping a pretrained feature extractor fixed and updating means and shared covariance statistics for a linear classifier as labeled examples arrive. On CORe50’s class-iid ordering, ResNet-18 Deep SLDA with online-updated covariance achieves Ω_all of 0.958, versus 0.873 for ExStream. This normalized trajectory score is not final accuracy; the feature extractor remains fixed.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.