Surviving the Data Deluge: A Combined Dynamical Systems/Machine Learning Approach
Abstract
This research sought to develop a comprehensive, computationally tractable framework for synthesizing information driven systems capable of both autonomously operating and supporting safety--critical human operations in rapidly changing ``data deluged' scenarios. Its conceptual backbone was a rigorous integration of systems theory, machine learning and optimization elements that emphasized robustness, computational simplicity and improved situational awareness. The research advanced the state of the art in systems theory by developing a tractable framework for robust identification/model (invalidation) of a broad class of dynamical systems that incorporates ideas from machine learning and semi-algebraic optimization to handle outliers, missing data and substantial noise levels.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 25, 2020
- Accession Number
- AD1104379
Entities
People
- Mario Sznaier
Organizations
- Northeastern University