{"ID":2857751,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2510.09825","arxiv_id":"2510.09825","title":"Decomposer Networks: Deep Component Analysis and Synthesis","abstract":"We propose the Decomposer Networks (DecompNet), a semantic autoencoder that factorizes an input into multiple interpretable components. Unlike classical autoencoders that compress an input into a single latent representation, the Decomposer Network maintains N parallel branches, each assigned a residual input defined as the original signal minus the reconstructions of all other branches. By unrolling a Gauss--Seidel style block-coordinate descent into a differentiable network, DecompNet enforce explicit competition among components, yielding parsimonious, semantically meaningful representations. We situate our model relative to linear decomposition methods (PCA, NMF), deep unrolled optimization, and object-centric architectures (MONet, IODINE, Slot Attention), and highlight its novelty as the first semantic autoencoder to implement an all-but-one residual update rule.","short_abstract":"We propose the Decomposer Networks (DecompNet), a semantic autoencoder that factorizes an input into multiple interpretable components. Unlike classical autoencoders that compress an input into a single latent representation, the Decomposer Network maintains N parallel branches, each assigned a residual input defined a...","url_abs":"https://arxiv.org/abs/2510.09825","url_pdf":"https://arxiv.org/pdf/2510.09825v1","authors":"[\"Mohsen Joneidi\"]","published":"2025-10-10T19:55:13Z","proceeding":"cs.LG","tasks":"[\"cs.LG\",\"cs.CV\",\"cs.IT\",\"cs.NE\"]","methods":"[]","has_code":false}