Skip to content

Latest commit

 

History

History
9 lines (5 loc) · 651 Bytes

File metadata and controls

9 lines (5 loc) · 651 Bytes

Effects of CPU and GPU architectures on the accuracy of Neural Networks

This repo is created for master thesis project "Effects of CPU and GPU architectures on the accuracy of Neural Networks". The thesis was written under the Data Science chair at the University of Passau. The repo contains the experiment to investigate the reason of variance in accuracy of image classification models when executed on different machines. The conclusion is CPUs, GPUs and GPU driver versions.

The exact experiment code is available at Zendo