Using random search for efficient hyper-parameters optimization with H2O
summary
Claims that random discrete search matches grid and manual search for model quality while using far less compute and cites the 2012 paper as evidence. Implements the feature inside H2O via the Python API using a search_criteria with max_models or max_runtime_secs plus early stopping, optimizes for metrics like AUC or MSE, runs on public datasets, produces the best-model parameters and exportable POJO artifacts, and the demos work end to end.
AI-written summary · about
chapters (8)
- 0:00H2O AI introduction
- 3:55Hyperparameter optimization need
- 5:49Optimization methods overview
- 8:48Random vs grid search
- 10:34Random search in H2O
- 14:32Demo: H2O Python API
- 17:46Random search benefits
- 19:28Q&A
related talks
44mScalable Automatic Machine Learning with H2O
30mScalable Automatic Machine Learning with H2O
01h 10mIntroduction to Machine Learning with H2O and Python
01h 10mGetting started with H2O on Python
18mBest Practices for Hyperparameter Optimization
26mModel Selection with Python: An Introduction to Hyper Parameter Tuning
01h 18mIntroduction to H20 AutoML with Python
50mScalable Data Science & Deep Learning with H2O