Applying the existing manifold-constrained hyper-connection (mHC) idea to speaker verification backbones yields small consistent EER reductions on VoxCeleb1 with no added parameters.
Datasets Training data.We used the V oxCeleb2 development set [28] (5,994 speakers) for training
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SD 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Beyond Residual Connections: Manifold-Constrained Hyper-Connections for Robust Speaker Representation Learning
Applying the existing manifold-constrained hyper-connection (mHC) idea to speaker verification backbones yields small consistent EER reductions on VoxCeleb1 with no added parameters.