A polynomial autoencoder beats PCA on transformer embeddings
A closed-form autoencoder: PCA encoder + quadratic Ridge decoder. On FiQA it gives 4× compression of mxbai-embed-large-v1 at -0.85 p.p. NDCG@10, +2.73 p.p. over PCA. No SGD, no neural networks.
Summary not yet generated for this article.
Comments (23 of ) ⌛