stareg-matlab: STructured Additive REGression in Matlab
Usage
rng(2); % set random seed
n_data = 500;
X = rand(n_data,2);
y = exp(-(X(:,1)-0.4).^2 ./ 0.01) + X(:,2).^2 + X(:,1).*X(:,2) + randn(n_data,1)*0.2; % create target data
Xtest = [rand(100,1), rand(100,1)]; % generate test data
ytest = exp(-(Xtest(:,1)-0.4).^2 ./ 0.01) + Xtest(:,2).^2 + Xtest(:,1).*Xtest(:,2);
d = {["s(1)", 100, "peak", 3000, "e"];
["t(1,2)", "12,20", "inc,inc", "2000,2000", "e,e"]};
[coef, basis_matrix, model] = Stareg.fit(d, X, y); % fit the model
ypred = Stareg.predict(Xtest, model, coef);
fprintf("MSE on test data = %4.4f \n", Utils.mse(ytest, ypred));Code Overview
Features
- Structured Additive Regression
- B-Splines, P-Splines and shape-constraint P-splines
- Incorporate prior knowledge using constraints:
- Monotonicity constraints: increasing, decreasing
- Shape constraints: convex, concave
- Peak and Valley constraints
- Interpretable Machine Learning
- Uni- and multivariate regression
Support
If you are having issues, please let us know. Contact can be made via GitHub.
License
The project is licensed under the MIT license.