Project 1 is about linear regression and the three main programs are: ''' regression.py results.py terrrain.py '''
This is a class consisting of the regression methods like OLS, ridge and lasso regression. It aslo contains a general train test split and k-fold cross validation where the number of folds can be chosen by the user. The main modules are the OLS, ridge, lasso and k_cross modules.
OLS(z = 2, X = 'None', test = False, full_matrices = False)
Ridge(lam, z = 2, X = 'None')
Lasso(lam = 1, z = 2, X ='None', max_iter=1001, precompute = False)
k_cross(X = 'None', z = 2, fold = 25, method2 = 'OLS', lam = 1, random_num = True, max_iter = 1001, precompute = False)
Other important modules are the 'z_tilde' module which uses the calculated beta and a given design matrix X to create the model. The remaining are:
z_tilde(beta, X = 'None')
MSE(z_tilde, z)
R_squared(z_tilde, z)
beta_variance(sigma_squared, X = 'None', lam = 0)
sigma_squared(sz_tilde, z, p = 'polynomial order')
lambda_best_fit(method, fold = 4, n_lambda = 1001, l_min = -5.5, l_max = -0.5, random_num = True, use_seed = False, seed = 42, X = 'None', z = 2, max_iter = 1001, full = False, precompute = True)
For better explanation see the documentation in the file.
This file creates all results from the Franke function data and saves the images in the Results folder.
This file creates all the results from the terrain data and saves the images in the Resutls_terrain folder.
results.py Results python file for the logreg results resutls.pyNN Results python file for the NN credit results regression_NN.py Results python file for NN Franke read_credit_and_preprocess.py Reading and preprocessing of the credit card data
reg_and_nn.py The main file containing the Logreg class, NN class and Gradient descent class Simple running example for the NN class
NN = NeuralNetwork(X_train, y_train, epochs = 200, n_cat = 1, eta = 10**(-4), batch_size = 500, end_activation = 'relu', split = False, cost_function = 'mse', tqdm_disable = True)
NN.add_layer(80, 'relu')
NN.initiate_network()
NN.train()
Project2_new is a backup if something is wrong with Project2 folder