CASL-VAE: Learning Structured Latent Variables from Unpaired Data for Semi-supervised Clustering and Paired Sample Generation

2026-07-09

AI Summary

This work addresses the challenge of disentangling multiple sources of variation in highly heterogeneous target populations under unpaired data settings, where existing methods struggle to separate distinct modes of specificity. The authors propose CASL-VAE, a deep contrastive latent variable model that decomposes variation into continuous shared factors across populations and target-specific hierarchical factors—combining discrete subtypes with continuous within-subtype variability to capture multi-level heterogeneity. Notably, CASL-VAE achieves joint cross-domain likelihood optimization without paired samples for the first time, enabling both semi-supervised clustering and paired sample generation. Experiments on semi-synthetic neuroimaging data demonstrate that CASL-VAE significantly outperforms baseline methods in subtype identification accuracy and generation quality, while revealing biologically plausible heterogeneity structures in Alzheimer’s disease.

Abstract

Quantifying variability in a target population relative to a reference population is central to many scientific and clinical problems (e.g., diseased vs. healthy). Yet, without paired data and in the presence of heterogeneous target variation, existing methods struggle to separate multiple modes of target-specific variation. We propose \textit{CASL-VAE}, a deep contrastive latent variable model that learns structured latent generative factors from unpaired data. CASL-VAE factorizes variation into continuous common latent factors shared across populations and hierarchical salient latent factors that model target-specific heterogeneity as discrete subtypes and continuous within-subtype variation. Using variational inference, we show how approximate joint likelihood optimization over reference and target domains can be performed using unpaired data, providing a principled basis for paired-sample generation and cross-domain analysis. We validate CASL-VAE on semi-synthetic neuroimaging data, demonstrating improved subtype recovery and paired-sample generation compared to baseline clustering and generative models. We also validate its ability to reveal biologically plausible heterogeneity in Alzheimer's disease.

Problem

• unpaired data
• target-specific variation
• heterogeneous variation
• semi-supervised clustering
• paired sample generation

Innovation

• structured latent variables
• unpaired data
• semi-supervised clustering
• paired sample generation
• hierarchical variation modeling