Missing Data in Signal Processing and Machine Learning: Models, Methods and Modern Approaches

TL;DR

This tutorial reviews various strategies for handling missing data in signal processing and machine learning, categorizing methods based on three analytical tasks, and discusses recent advances and future research directions.

Contribution

It introduces a unified conceptual framework for missing data strategies, emphasizing practical case studies and encouraging development of new methods for real-world applications.

Findings

Case studies demonstrate effectiveness of imputation techniques

Identification of promising research directions

Comparison of methods across different applications

Abstract

This tutorial aims to provide signal processing (SP) and machine learning (ML) practitioners with vital tools, in an accessible way, to answer the question: How to deal with missing data? There are many strategies to handle incomplete signals. In this paper, we propose to group these strategies based on three common analytical tasks: i) missing-data imputation, ii) estimation with missing values and iii) prediction with missing values. We focus on methodological and experimental results through specific case studies on real-world applications. Promising and future research directions are also discussed. We hope that the proposed conceptual framework and the presentation of recent missing-data problems related will encourage researchers of the SP and ML communities to develop original methods and to efficiently deal with new applications involving missing data.