The aim of this page is to point to various resources at NIH related to the fields of machine learning and deep learning.


Biowulf

NIH investigators interested in running machine learning and deep learning workflows have access to the Biowulf cluster.  The cluster has many preinstalled software packages covering classical machine learning algorithms in Python, R, etc. (e.g., scikit-learn).  Similarly, deep learning frameworks (e.g., TensorFlow, Caffe, Keras, etc.) are available on Biowulf, as discussed here.


CANDLE

Training machine learning models consists of parameter and hyper-parameter optimizations and large amounts of data processing, all of which require significant compute resources.  The National Cancer Institute and the Department of Energy developed the CANcer Distributed Learning Environment (CANDLE) software in order to scale machine and deep learning frameworks on High Performance Computing (HPC) clusters.

To get started, here is a tutorial: 

https://cbiit.github.io/sdsi/candle/

Contacts: Andrew WeismanGeorge Zaki

Created by Andrew Weisman Last Modified Tue February 18, 2020 10:06 am by George Zaki