How to Design Multiplierless Neural Networks for Deep Learning?

EasyChair Preprint no. 13105

Abstract

The paper is an extended version of the paper accepted at MECO2024 with a procedure for designing neural networks.  The accepted paper describes the efficient implementation of neural networks based on symbolic analysis. The main advantage is the reduction of processing latency by replacing general-purpose multipliers with a small number of summing components.

Keyphrases: deep learning, neural networks, probability, sensitivity, Symbolic Analysis

@Booklet{EasyChair:13105,
  author = {Maja Lutovac Banduka and Miroslav Lutovac},
  title = {How to Design Multiplierless Neural Networks for Deep Learning?},
  howpublished = {EasyChair Preprint no. 13105},

  year = {EasyChair, 2024}}