Multimodal Deep Generative Model for Semi-Supervised Learning under Class Imbalance
Heegeon Yoon, Heeyoung Kim · 2026
Read Multimodal Deep Generative Model for Semi-Supervised Learning under Class Imbalance by Heegeon Yoon, Heeyoung Kim (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-06289.txt.
Description
Heegeon Yoon and Heeyoung Kim build a multimodal generative model for learning from partly labeled, class imbalanced data. Models trained on skewed data favor majority classes, and semi supervised learning makes it worse: pseudo labels for unlabeled items are predicted from the same skewed labeled set, spreading the bias further. Most existing remedies also assume a single input modality, ignoring the complementary signals that multimodal data offers.
Their model uses a separate encoder per modality with latent variables shared across modalities, joined by a product of experts rule. Student t distributions replace Gaussians in the prior, encoder, and decoder to fit the heavy tailed latent patterns of imbalanced data. A new training objective built on a power divergence covers labeled and unlabeled data together. Benchmark results beat baseline methods.
Catalog record
| CREATOR | Heegeon Yoon, Heeyoung Kim |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 62 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.06289