ML model tuning with constraints (CNN tuning for MNIST example)
Welcome to video #4 in the Adaptive Experimentation series, presented by graduate student Sterling Baird @sterling-baird at the 18th IEEE Conference on eScience in Salt Lake City, UT (October 10-14, 2022). In this video, Sterling demonstrates how to use Ax to optimize machine learning models with constraints by tuning a convolutional neural network to classify digits in the MNIST dataset. In the next video of the series, we will explore batch optimization of expensive functions, such as simulations. Stay tuned!
Github link to jupyter notebook https://github.com/sparks-baird/self-...
previous video in series: • Closed-loop optimization of inexpensi...
next video in series: • Batch optimization of expensive funct...
0:00 MNIST dataset and task
0:28 libraries, data, and function set up
2:25 run optimization loop
3:12 examining results
3:44 classification performance vs iteration