ZAYAN: Disentangled Contrastive Transformer for Tabular Remote Sensing Data

arXiv cs.LG / 5/1/2026

📰 NewsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper introduces ZAYAN, a self-supervised, feature-centric contrastive learning framework designed for tabular data in remote sensing and environmental science where data heterogeneity and redundant features are common.
ZAYAN contrasts features rather than samples, eliminating the need for explicit anchor selection and avoiding reliance on class labels while targeting a redundancy-minimized, disentangled embedding space.
The approach consists of two modules: ZAYAN-CL for zero-anchor pretraining using dynamic perturbations and masking, and ZAYAN-T, a Transformer that leverages the learned feature embeddings for downstream classification.
Experiments on eight datasets (six remote-sensing tabular benchmarks plus two flood-prediction tables) show ZAYAN delivers better accuracy, robustness, and generalization than tabular deep learning baselines, with consistent improvements under label scarcity and distribution shift.

Abstract

Learning informative representations from tabular data in remote sensing and environmental science is challenging due to heterogeneity, scarce labels, and redundancy among features. We present ZAYAN (Zero-Anchor dYnamic feAture eNcoding), a self-supervised, feature-centric contrastive framework for tabular data. ZAYAN performs contrastive learning at the feature rather than sample level, removing the need for explicit anchor selection and any reliance on class labels, while encouraging a redundancy-minimized, disentangled embedding space. The framework has two modules: ZAYAN-CL, which pretrains feature embeddings via a zero-anchor contrastive objective with dynamic perturbations and masking, and ZAYAN-T, a Transformer that conditions on these embeddings for downstream classification. Across eight datasets, including six remote-sensing tabular benchmarks and two remote-sensing-driven flood-prediction tables from satellite and GIS products, ZAYAN achieves superior accuracy, robustness, and generalization over tabular deep learning baselines, with consistent gains under label scarcity and distribution shift. These results indicate that feature-level contrastive learning and dynamic feature encoding provide an effective recipe for learning from tabular sensing data.