Contrastive Dimension Reduction: A Systematic Review.

Area of Science:

• Data Science
• Statistical Learning
• Manifold Learning
• Dimension Reduction
• High Dimensional Data Analysis

Background:

• Traditional dimension reduction techniques like PCA may not effectively isolate treatment-specific signals.
• Contrastive dimension reduction (CDR) methods are designed to extract unique signals from a treatment group compared to a control group.
• This problem is prevalent in genomics, imaging, and time series analysis.

Purpose of the Study:

• To provide a systematic overview of existing Contrastive Dimension Reduction (CDR) methods.
• To propose a unified conceptual framework and taxonomy for CDR techniques.
• To identify key applications, challenges, and future research directions in CDR.

Main Methods:

• Systematic review of existing Contrastive Dimension Reduction (CDR) literature.
• Development of a pipeline for analyzing case-control studies using CDR.
• Classification of CDR methods based on assumptions, objectives, and mathematical formulations.

Main Results:

• A comprehensive taxonomy unifying disparate CDR approaches under a shared conceptual framework.
• Identification of key applications and current challenges associated with CDR methods.
• A proposed pipeline to facilitate the analysis of case-control studies.

Conclusions:

• CDR methods offer a powerful approach to extracting treatment-specific signals in high-dimensional data.
• A unified framework and taxonomy are crucial for broader adoption and further development of CDR.
• Further research is needed to address existing challenges and explore new frontiers in CDR.