Published on:
10 October 2023
Primary Category:
Distributed, Parallel, and Cluster Computing
Paper Authors:
Lakshmi Arunachalam,
Fahim Mohammad,
Vrushabh H. Sanghavi
Transfer learning builds on existing knowledge from pre-trained models
Case study classifies cancer tissue with 94.5% accuracy using ResNet on TensorFlow
Intel Xeon CPUs challenge GPU-centric training mindset
Mixed precision with Intel AMX and Horovod distribution optimize performance
Transfer learning for image classification
This paper explores transfer learning for image classification, showing how pre-trained models can enable high accuracy with minimal training time and resources. The authors demonstrate a case study classifying cancer tissue types, achieving 94.5% accuracy using ResNet and TensorFlow on Intel Xeon CPUs. They highlight techniques like mixed precision with Intel AMX and distributed training with Horovod to optimize performance.
No comments yet, be the first to start the conversation...
Sign up to comment on this paper