Evolving Deep Networks Using HPC

By Steven R. Young, Derek C. Rose, J. Travis Johnston, William T. Heller, T. Karnowski, Thomas E. Potok, Robert M. Patton, Gabriel N. Perdue, Jonathan Miller. 2017

45 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Tal Ben-Nun
2. Torsten Hoefler
3. Lei Shao
4. Shirley Ho
5. Siyu He
6. Peter Richtárik
7. Ahmed Khaled
8. Dan Alistarh

In Computer Vision and Pattern Recognition

1. Torsten Hoefler
2. Tal Ben-Nun
3. Shirley Ho
4. Peter Richtárik
5. Ahmed Khaled
6. Dan Alistarh

Citations in the corpus

• Decentralized gradient methods: does topology matter? (2020, neutral): “For example, experiments in (Young et al., 2017) rely on more than 10 4 computing nodes to iteratively improve the (hyper)parameters of a deep neural network.”
• Demystifying Parallel and Distributed Deep Learning (2018, methodology): “GAs are highly amenable to parallelism, and have been successfully used for very large-scale training [261], where 18,000 nodes were used on the Titan supercomputer for 24h to obtain state-of-the-art accuracy for segmentation and reconstruction problems.”
• CosmoFlow: Using Deep Learning to Learn the Universe at Scale (2018, methodology): “For example, Young et al. (2017) [17] scanned the possible space of network hyperparameters, using up to 18,000 GPUs.”