Continuous benchmark suite for the scikit-learn project.

In order to run the benchmarks on your own machine, please follow these steps.

1. Clone the repository somewhere, for example ~/code/scikit-learn-speed

2. Extract the datasets:

   cd ~/code/scikit-learn-speed/benchmarks
   tar jxvf data.tar.bz2
   

3. Create the configuration file ~/.vbench-skl. For example:

   [setup]
   repo_path = /Users/vene/code/scikit-learn
   repo_url = git@github.com:scikit-learn/scikit-learn.git
   db_path = /Users/vene/code/scikit-learn-speed/benchmarks/benchmarks.db
   tmp_dir = /tmp/vb_sklearn
   

The values displayed above are hardcoded defaults, and they are used in case the configuration value doesn't exist, or to override skipped values. Specifically, this means you don't have to bother to set repo_url and tmp_dir.

4. From the scikit-learn-speed/benchmarks folder, run:

   python run_suite.py  # This runs the entire suite, ~10min on my machine
   python generate_rst_files.py  # This prepares the rst documentation
   

5. To actually generate the HTML files, change to the scikit-learn-speed foldar and execute::

   python make.py
   

6. You can view the results by opening scikit-learn-speed/benchmarks/build/html/index.html.

The following datasets are available:

• arcene: train: (100, 10000), test: (100, 10000)
• madelon: train: (2000, 500), test: (600, 500)
• minimadelon: train: (30, 500), test: (20, 500), 10 output
• blobs: train: (300, 50) test: (200, 50), 10 tight centers
• newsgroups: sparse, train: (11214, 130088), test: (7432, 130088)

In addition, you can append the following options to any dataset's name:

• -oney: Only keeps the first output, i. e. y = y[:, 0]. Necessary for estimators that don't support multidimensional output arrays.
• -semi: Unlabels samples at random, by setting the corresponding output to -1. Useful for semi-supervised algorithms.