Welcome to ROLCH

ROLCH is a Python package for online distributional learning and online models for conditionally heteroskedastic data. We provide an online implementation of the well-known GAMLSS model using online coordinate descent (OCD).

Note

ROLCH is currently in the first alpha phase. Please expect changes to happen frequently.

Installation

ROLCH is available on the Python Package Index and can be installed via pip:

pip install rolch

Example

The following few lines give an introduction. We use the diabetes data set and model the response variable \(Y\) as Student-\(t\) distributed, where all distribution parameters (location, scale and tail) are modelled conditional on the explanatory variables in \(X\).

import rolch
import numpy as np
from sklearn.datasets import load_diabetes

np.set_printoptions(precision=3, suppress=True)

# Load Diabetes data set
# Add intercept (will not be regularized)
X, y = load_diabetes(return_X_y=True)
X = np.hstack((np.ones((X.shape[0], 1)), X))

dist = rolch.DistributionT()

# Initialise the estimator
online_gamlss_lasso = rolch.OnlineGamlss(
    distribution=dist, 
    method="lasso", 
    estimation_kwargs={"ic" : {i: "bic" for i in range(dist.n_params)}}
)

# Fit the model and print coefficients
# We fit on all but the last data point
online_gamlss_lasso.fit(
    y[:-1], 
    X[:-1, :], 
    X[:-1, :], 
    X[:-1, :]
)

print("LASSO Coefficients \n")
print(np.vstack(online_gamlss_lasso.betas).T)

# Update the fit and print new coefficients
online_gamlss_lasso.update(
    y[[-1]], 
    X[[-1], :], 
    X[[-1], :], 
    X[[-1], :]
)

print("\nCoefficients after update call \n")
print(np.vstack(online_gamlss_lasso.betas).T)