Binary Forward-Only Algorithms

Baichuan Huang, Amir Aminifar

Abstract: Editor’s notes: This article presents a detailed evaluation of binarization strategies in state-of-the-art forward-only learning algorithms. This work provides valuable insights into the practical boundaries of efficient learning under forward-only paradigms and will be of strong interest to researchers in ondevice learning TinyML systems. —Theocharis Theocharides, University of Cyprus, Cyprus —Marian Verhelst, KU Leuven, Belgium —Vijay Janapa Reddy, Harvard University, USA —Evgeni Gousev, Qualcomm, USA

External IDs:doi:10.1109/mdat.2025.3528366