Approximately Quantizing Algorithm for In-memory Machine Learning Classifier

摘要

This work proposed Approximate Quantizing, a simple 1-Bitrnquantization algorithm for after trained linear model weight. Thernproposed method can reach the similar accuracy as original 64-rnBit in MNIST database (～91%) and with faster (<7% of time)rnand lower computing complexity (<25% of columns needed),rncompared to conventional CRR algorithm.