Hyper Parameter Search using Optuna
Designing Neural Network learning algorithm requires setting many hyper parameters. In this 2 hour workshop we will see how we can use the Optuna python package to automate the labourious task of finding good hyper parameters.
Overview
The workshop is organized around introductory material giving some background to what hyper parameter optimization is about and concepts underlying Bayesian Optimization. The first practical part uses a Jupyter Notebook to illustrate how the Optuna package works, and highlights that it allows us to essentially optimize any black box.
The second practical shows how we can add hyper parameter optimization to a simple neural network trained using cross-validation to automate the search.
Learn More
Visit the lesson page to access the full content.