Multi-modal Differentiable Unsupervised Feature Selection

Junchen YangOfir LindenbaumYuval KlugerAriel Jaffe

Mar 2023

摘要

Multi-modal high throughput biological data presents a great scientificopportunity and a significant computational challenge. In multi-modalmeasurements, every sample is observed simultaneously by two or more sets ofsensors. In such settings, many observed variables in both modalities are oftennuisance and do not carry information about the phenomenon of interest. Here,we propose a multi-modal unsupervised feature selection framework: identifyinginformative variables based on coupled high-dimensional measurements. Ourmethod is designed to identify features associated with two types of latentlow-dimensional structures: (i) shared structures that govern the observationsin both modalities and (ii) differential structures that appear in only onemodality. To that end, we propose two Laplacian-based scoring operators. Weincorporate the scores with differentiable gates that mask nuisance featuresand enhance the accuracy of the structure captured by the graph Laplacian. Theperformance of the new scheme is illustrated using synthetic and real datasets,including an extended biological application to single-cell multi-omics.

展开全部

图表提取

论文十问

参考文献

被引用

提问与回答

暂无人提供速读十问回答

被引用

笔记

问答

领域