TY - JOUR
T1 - Differential Geometric View of Information Flow in Neural Nets
AU - Sreehari, Suhas
AU - Ramuhalli, Pradeep
AU - Liu, Frank
N1 - Publisher Copyright:
© 2025, Society for Imaging Science and Technology.
PY - 2025
Y1 - 2025
N2 - In this paper, we explore a space-time geometric view of signal representation in machine learning models. The question we are interested in is if we can identify what is causing signal representation errors – training data inadequacies, model insufficiencies, or both. Loosely expressed, this problem is stylistically similar to blind deconvolution problems. However, studies of space-time geometries might be able to partially solve this problem by considering the curvature produced by mass in (Anti-)de Sitter space. We study the effectiveness of our approach on the MNIST dataset.
AB - In this paper, we explore a space-time geometric view of signal representation in machine learning models. The question we are interested in is if we can identify what is causing signal representation errors – training data inadequacies, model insufficiencies, or both. Loosely expressed, this problem is stylistically similar to blind deconvolution problems. However, studies of space-time geometries might be able to partially solve this problem by considering the curvature produced by mass in (Anti-)de Sitter space. We study the effectiveness of our approach on the MNIST dataset.
UR - https://www.scopus.com/pages/publications/105019037456
U2 - 10.2352/EI.2025.37.14.COIMG-141
DO - 10.2352/EI.2025.37.14.COIMG-141
M3 - Conference article
AN - SCOPUS:105019037456
SN - 2470-1173
VL - 37
JO - IS and T International Symposium on Electronic Imaging Science and Technology
JF - IS and T International Symposium on Electronic Imaging Science and Technology
IS - 14
M1 - 141
T2 - IS and T International Symposium on Electronic Imaging 2025: 23rd Computational Imaging, COIMNG 2025
Y2 - 2 February 2025 through 6 February 2025
ER -