arXiv AI

No Data Wasted: A Semi-supervised Generative Model for Incomplete Multi-view Data Integration with Missing Labels

The paper presents a semi‑supervised generative model for multi‑view learning that handles missing views and missing labels. It combines a likelihood‑based approach for unlabeled data with an information bottleneck (IB) framework for labeled data, incorporating modality‑specific information and cross‑view mutual information maximization to learn a shared latent space. Experiments show improved predictive and generative performance on complex datasets with limited labeled samples.

arXiv Machine Learning
Aug 20

Pretraining Reusable Inference Across Views with Synthetic Task Priors

The paper introduces SIMPLE, a prior‑fitted multi‑view in‑context learner that learns a reusable, task‑conditioned inference procedure instead of a fixed fusion function. By generating synthetic task priors in embedding space, SIMPLE can handle diverse view configurations, class structures, and missingness patterns. Experiments on multi‑view and multi‑omics benchmarks show that a frozen SIMPLE model performs competitively, and lightweight adapter calibration further improves performance across most datasets.

By Jielong Lu, Zhihao Wu, Jiajun Yu, Zhaoliang Chen, Haishuai Wang
arXiv Computer Vision
Nov 5

Semantic Prior-Guided Incomplete Cross-Modal Retrieval with Noisy Labels

The paper introduces SPIRN, a semantic prior-guided framework for incomplete cross‑modal retrieval that simultaneously handles modality incompleteness and noisy labels. SPIRN employs fixed semantic priors to create a stable category‑level space and incorporates modules for quality‑aware robust learning, global label calibration, and confidence‑driven selective imputation. Experiments on four benchmark datasets show that SPIRN outperforms 11 state‑of‑the‑art methods across scenarios with missing modalities and noisy annotations.

By Jiali Chen, Ruitao Pu, Yingke Chen, Dezhong Peng
arXiv Machine Learning
Oct 9

Marformer: A Transformer for Predicting Missing Data Distributions

The Marformer is a Transformer model designed to predict conditional marginal distributions of missing variables given observed data, enabling computation of Bayes risk and Value of Information. It learns a hidden-vector representation for each variable’s distribution and refines these through attention, similar to BERT’s masked language modeling but for probabilistic inference. Evaluations on synthetic Bayesian networks, discretized Gaussians, structured annotation data, and a real annotation dataset show that the Marformer matches or surpasses classical missing-data methods while being faster and not requiring explicit joint distribution modeling.

By Prabhav Singh, Xiheng Tom Wang, Haojun Shi, Jason Eisner