Computational Paradox of Deep Learning: A Qualitative Explanation

Authors

Jonatan Contreras, The University of Texas at El PasoFollow
Martine Ceberio, The University of Texas at El PasoFollow
Olga Kosheleva, The University of Texas at El PasoFollow
Vladik Kreinovich, The University of Texas at El PasoFollow
Nguyen Hoang Phuong, Thang Long UniversityFollow

Publication Date

6-1-2022

Comments

Technical Report: UTEP-CS-22-74

Abstract

In general, the more unknowns in a problem, the more computational efforts is necessary to find all these unknowns. Interestingly, in state-of-the-art machine learning methods like deep learning, computations become easier when we increase the number of unknown parameters way beyond the number of equations. In this paper, we provide a qualitative explanation for this computational paradox.