Algorithm-Hardware Co-Design of Distribution-Aware Logarithmic-Posit Encodings for Efficient DNN Inference

arXiv:2403.05465v1 Announce Type: cross
Abstract: Traditional Deep Neural Network (DNN) quantization methods using integer, fixed-point, or floating-point data types struggle to capture diverse DNN parameter distributions at low precision, and often require large silicon overhead and intensive quantization-aware training. In this study, we introduce Logarithmic Posits (LP), an adaptive, hardware-friendly data type inspired by posits that dynamically adapts to DNN weight/activation distributions by parameterizing LP bit fields. We also develop a novel genetic-algorithm based framework, LP Quantization (LPQ), to find …

abstract algorithm arxiv cs.ai cs.ar cs.lg cs.ne data deep neural network design distribution diverse dnn fixed-point hardware inference low network neural network posit precision quantization silicon struggle study training type types