benoitdescamps / Hyperparameters-tuning

Some notes/codes on hyperparameters tuning techniques with some hacking around...

Hyperparameters-tuning

Series of notes/code-snippets about various hyperparameter optimisation techniques.

Allocation-based optimisations

SuccessiveHalving

Code is based on the paper by Talwalkar A. Non-stochastic best arm identification and hyperparameter optimization. Some Example code for :

• sklearn: RandomForest, GradientBoosting, etc...
• xgboost

More information can be found in this blog post

Hyperband

Code is based on the paper by Talwalkar A. Hyperband: A Novel Bandit-Based Approach to Hyperparameter Optimization.

TODO:

• add setup.py
• add example Extreme-random-forest
• add proper hyperparameter-spaces in examples
• ? comparison for some dataset?

References:

• zygmuntz hyperband

Random Search

Spark implementation of Random Search for sampling over breeze.stats.distribution. More information can be found in this blog post