{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/autoencoder-node-saliency-selecting-relevant","title":"Autoencoder Node Saliency: Selecting Relevant Latent Representations","arxiv_id":"1711.07871","date":"2017-11-21","proceeding":null,"authors":["Ya Ju Fan"],"abstract":"The autoencoder is an artificial neural network model that learns hidden\nrepresentations of unlabeled data. With a linear transfer function it is\nsimilar to the principal component analysis (PCA). While both methods use\nweight vectors for linear transformations, the autoencoder does not come with\nany indication similar to the eigenvalues in PCA that are paired with the\neigenvectors. We propose a novel supervised node saliency (SNS) method that\nranks the hidden nodes by comparing class distributions of latent\nrepresentations against a fixed reference distribution. The latent\nrepresentations of a hidden node can be described using a one-dimensional\nhistogram. We apply normalized entropy difference (NED) to measure the\n\"interestingness\" of the histograms, and conclude a property for NED values to\nidentify a good classifying node. By applying our methods to real data sets, we\ndemonstrate the ability of SNS to explain what the trained autoencoders have\nlearned.","url_abs":"http://arxiv.org/abs/1711.07871v2","url_pdf":"http://arxiv.org/pdf/1711.07871v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"autoencoder-node-saliency-selecting-relevant","repo_url":"https://github.com/LLNL/ANS","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[],"methods":[{"method_slug":"pca","method_name":"PCA"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}