cyc-gbm 0.0.24

A python package for the Cyclical Gradient Boosting Machine algorithm

cyc-gbm

A package for the Cyclical Gradient Boosting Machine algorithm. For the (pre-print) paper describing the algorithm, see here.

Installation

You can install the package using pip:

pip install cyc-gbm

Alternatively, you can install the package from source. This will also include scripts for reproducing the results in the papar. Follow these steps:

1. Clone this repository to your local machine:

   git clone https://github.com/henningzakrisson/c-gbm.git
   

2. Create a virtual environment in the root directory of the repository:

   python3 -m venv venv
   

3. Activate the virtual environment:

   source venv/bin/activate
   

4. Install the required dependencies:

   pip install -r requirements.txt
   

Usage example

Fitting the mean and (log) sigma parameters of a normal distribution to a simulated dataset:

import numpy as np
from cyc_gbm import CyclicalGradientBooster
from sklearn.model_selection import train_test_split

# Simulate data
X = np.random.normal(size=(1000, 2))
mu = X[:, 0] + 10 * (X[:, 1] > 0)
sigma = np.exp(3 - 2 * (X[:, 0] > 0))
y = np.random.normal(mu, sigma)

# Split data
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)

# Fit model
model = CyclicalGradientBooster(
   distribution='normal',
   kappa=[26, 34],
   eps=0.1,
   max_depth=2,
   min_samples_leaf=20,
)
model.fit(X_train, y_train)

# Evaluate
loss = model.dist.loss(y=y_test, z=model.predict(X_test)).sum()
print(f'negative log likelihood: {loss}')

Reproducing the numerical illustrations in the paper

The numerical illustrations in the paper can be reproduced by running the numerical_illustration function in the numerical_illustration/numerical_illustration.py module. The function takes the path to a configuration file as input. The configuration file is a yaml file that specifies the parameters of the numerical illustration. An example configuration file can be found in numerical_illustration/config/simulation_config.yaml. For running several experiments in one run, I refer to the numerical_illustrations function in the same module. See the documentation for usage. An example configuration file for running several experiments can be found in numerical_illustration/config/simulation_run/master_config.yaml.

Not yet implemented

• Add support for categorical features (currently the trees are based on sklearn.tree.DecisionTreeRegressor which does not support categorical features)
• Add other tuning methods (such as adaptive shrinkage)

Contact

If you have any questions, feel free to contact me here.