Joint Identifiability of Cross-Domain Recommendation via Hierarchical Subspace Disentanglement

arXiv:2404.04481v1 Announce Type: cross
Abstract: Cross-Domain Recommendation (CDR) seeks to enable effective knowledge transfer across domains. Existing works rely on either representation alignment or transformation bridges, but they struggle on identifying domain-shared from domain-specific latent factors. Specifically, while CDR describes user representations as a joint distribution over two domains, these methods fail to account for its joint identifiability as they primarily fixate on the marginal distribution within a particular domain. Such a failure may overlook the conditionality between two domains …

abstract alignment arxiv cs.ai cs.ir cs.lg distribution domain domains hierarchical knowledge recommendation representation struggle transfer transformation type via