Lisbon Data Science Meetup - Machine Learning Hyper Parameter Search (from Grid Search to SKOPT)
Wednesday 19 October 2016, 20:00 - 21:00
Rua da Prata, 80, Lisbon, Portugal
An important part of using any machine learning algorithm (such as Logistic Regression or Random Forests) is tuning the hyper parameters. Tools such as scikit's Grid Search and Randomized Search have greatly improved the ability to use CPU instead of sweat and tears to tune large hyper-parameter spaces. More recently, a new library has joined the toolkit: SKOPT, which allows us to use Bayesian Optimization to find the best hyper-parameters faster, and better. On the 19th, Hugo Lopes (Data Scientist @CrowdProcess) has generously agreed to go over all of these methods, and to help de-mystify the all powerful SKOPT. As always, please reserve your place only if you are certain that you can come, as spaces are limited.
Publicado por: Betabeers