Reproduce “Flexible model composition in machine learning and its implementation in MLJ” 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
Anthony D. Blaom, Sebastian J. Vollmer
Blaom and Vollmer propose learning networks, graph-based compositions that separate model settings, learned parameters, and operations. The design lets different operations share one machine, while separate training events use the same model settings. The authors illustrate this with learned target transformations and stacking using out-of-sample base predictions. These examples show expressibility in MLJ, not improved predictive performance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.