Time series representation learning in Health area

• Grant number: 22/00301-0
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Effective date (Start): April 01, 2022
• Effective date (End): March 31, 2023
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Diego Furtado Silva
• Grantee: Sophia Santonastasio Schuster
• Host Institution: Centro de Ciências Exatas e de Tecnologia (CCET). Universidade Federal de São Carlos (UFSCAR). São Carlos , SP, Brazil

Abstract

With the large volume of data collected continuously and on a daily basis in several applications, it is increasingly common to obtain and store time series without any type of annotation. The challenge of labeling such data is due to the high cost of manually aggregating any information to the collected data. This situation becomes even more significant in the domain of healthcare applications, where there are several ways to obtain temporal data. The Creation of models from data in this domain can have a real impact on the lives of many people. On the other hand, identifying and annotating any aspect of this scope is highly specialized and therefore expensive work. In this scenario, this research aims to investigate self-supervised learning techniques based on data augmentation that can be applied in the time series domain, with a special interest in health-related applications. This is a challenging and unprecedented strategy in tasks involving time series, in addition to the probability of having a high impact on the chosen application domain. (AU)