Global ETD Search

Comparing the Performance of Small Word-Size Floating-Point Numerics to Fixed-Point Numerics in Neural Networks

Vennapusa, Lakshmi GrishmaGeorgia Institute of Technology2025

… reduced-precision floating-point formats (FP8, FP12, FP16) for dynamic, layer-wise quantization during training. APT monitors per-layer Quantization Error Measurement (QEM) to guide precision adjustments and incorporates a novel bit-shuffling mechanism to reallocate bits between exponent …

gatech Repository record for Comparing the Performance of Small Word-Size Floating-Point Numerics to Fixed-Point Numerics in Neural Networks (opens in a new tab)