InfoDPCCA combines a dynamic probabilistic CCA model with an information-bottleneck objective so the shared latent state is trained to contain only the mutual information of the two sequences and still predict the next observations.
Identity mappings in deep residual networks
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
InfoDPCCA: Information-Theoretic Dynamic Probabilistic Canonical Correlation Analysis
InfoDPCCA combines a dynamic probabilistic CCA model with an information-bottleneck objective so the shared latent state is trained to contain only the mutual information of the two sequences and still predict the next observations.