Skip to main content

Hyperparameter Tuning

Natalia Ponomareva ( Google NY )

Hyperparameter tuning is key area in machine learning and is an active area of research. Machine learning algorithms almost always contain hyperparameters (a couple to a couple dozen parameters that control the model complexity and training behavior) which have significant impact on model performance and therefore need to be tuned. Searching the parameter space in a brute force manner can be prohibitively costly. Algorithm-assisted hyperparameter tuning addresses this. This is an active area of research which looks for the methods that will find the best values of hyperparameters automatically and as fast as possible. In this presentation I will talk about Bayesian hyperparameter tuning algorithms, their types and characteristics and briefly mention the latest research in hyperparameter tuning.

Speaker bio

Natalia Ponomareva is a software engineer with more than 9 years of experience and a career that spans three countries. She holds a Master's degree in Applied Math from a Russian university and a Master's degree in Computer science from Oxford University. In her pre-Google life she had a chance to work for a software outsourcing company, a financial company and a media startup. Currently Natalia works in Research, implementing and applying machine learning algorithms to various tasks at Google.

 

 

Share this: