A joint prior that aligns modality-specific GP kernels with the latent similarity matrix improves cross-modal retrieval for GPLVM-based models on four benchmarks.
On the role of correlation and abstraction in cross-modal multimedia retrieval,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2019 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Harmonized Multimodal Learning with Gaussian Process Latent Variable Models
A joint prior that aligns modality-specific GP kernels with the latent similarity matrix improves cross-modal retrieval for GPLVM-based models on four benchmarks.