Reproduce “Feature space reduction as data preprocessing for the anomaly detection” 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
Simon Bilik, Karel Horak
The paper tests whether reconstruction errors can form a compact feature space for detecting faulty Tarallini biscuits when training uses only acceptable images. Convolutional autoencoders reconstruct images, and L2 and SSIM errors become two features, compared with PCA plus t-SNE embeddings. On the Cookie Dataset, this separated OK and NOK more clearly, but clusters overlapped and OC-SVM underperformed.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.