Beyond Grid and Random Search: Tuning Deep Learning Pipelines With Genetic Algorithms
Expand access to the essential black-box optimization technique and the range of its potential applications.
About the webinar
How many hidden units should be used in your neural network? What is the ideal learning rate for your data? How about the dropout percentage, and on which layers?
These are important questions for any deep learning practitioner.
Developed by SAS and the new solveBlackbox action, autotuning can help you find the ideal set of hyperparameters for a chosen modeling algorithm.
It gives you the flexibility to use a wide array of customized programs and solves the most challenging optimization problems.
Join our expert presenters as they share how to:
- Apply autotuning to your deep learning models with genetic algorithms.
- Modify the structure of a convolutional neural network for object detection tasks.
- Use cache in and cache out tables for repetitive searches.