{"paper":{"title":"Dataset-Driven Channel Masks in Transformers for Multivariate Time Series","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Channel masks from similarity matrices and learnable domain parameters enable partial channel dependence in Transformer attention for multivariate time series.","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Kibok Lee, Seunghan Lee, Taeyoung Park","submitted_at":"2024-10-30T17:12:03Z","abstract_excerpt":"Recent advancements in foundation models have been successfully extended to the time series (TS) domain, facilitated by the emergence of large-scale TS datasets. However, previous efforts have primarily Capturing channel dependency (CD) is essential for modeling multivariate time series (TS), and attention-based methods have been widely employed for this purpose. Nonetheless, these methods primarily focus on modifying the architecture, often neglecting the importance of dataset-specific characteristics. In this work, we introduce the concept of partial channel dependence (PCD) to enhance CD mo"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Channel masks consisting of a similarity matrix and dataset-specific learnable domain parameters, integrated via element-wise multiplication into attention matrices, achieve partial channel dependence and thereby enhance channel dependency modeling in Transformer-based multivariate time series models.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That a similarity matrix derived from the data plus a modest number of learnable domain parameters will reliably isolate the relevant partial dependencies without introducing harmful bias or requiring per-dataset hyper-parameter search that negates the claimed benefit.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Introduces channel masks built from similarity matrices plus learnable domain parameters to realize partial channel dependence inside Transformer attention for multivariate time series.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Channel masks from similarity matrices and learnable domain parameters enable partial channel dependence in Transformer attention for multivariate time series.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"6e8778362d05f2fb69bec99abb75fa5ae6eb7552f46ea310301d054547c224b4"},"source":{"id":"2410.23222","kind":"arxiv","version":4},"verdict":{"id":"7e77a271-f676-4fd9-8770-914fd23927b4","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-23T18:36:12.791850Z","strongest_claim":"Channel masks consisting of a similarity matrix and dataset-specific learnable domain parameters, integrated via element-wise multiplication into attention matrices, achieve partial channel dependence and thereby enhance channel dependency modeling in Transformer-based multivariate time series models.","one_line_summary":"Introduces channel masks built from similarity matrices plus learnable domain parameters to realize partial channel dependence inside Transformer attention for multivariate time series.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That a similarity matrix derived from the data plus a modest number of learnable domain parameters will reliably isolate the relevant partial dependencies without introducing harmful bias or requiring per-dataset hyper-parameter search that negates the claimed benefit.","pith_extraction_headline":"Channel masks from similarity matrices and learnable domain parameters enable partial channel dependence in Transformer attention for multivariate time series."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2410.23222/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}