Reproduce “Contextual Unsupervised Outlier Detection in Sequences” 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
Mohamed A. Zahran, Leonardo Teixeira, Vinayak Rao, Bruno Ribeiro
The authors present a method for finding unusual actions in user sequences when labels are scarce. It ranks actions against replacements, adjusts for frequency, and calibrates a threshold to a chosen false-positive rate using sequences from a fitted normal model. On synthetic Last.fm data, SwitchingFlow outperforms tested methods across almost all rates; this is not validation on naturally occurring anomalies.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.