Numpy implementations of basic ML algorithms.

Currently implemented:

• Linear regression (least squares)
• Bayesian linear regression
• Logistic regression
• Neural networks

Also includes polynomial and radial basis functions.

TODO:

• k-means
• k-NN (+ PCA for low dimensional embedding before k-NN)
• Gaussian processes (with basic covariance functions like squared exponential).

regressor = LeastSquaresRegressor(scalar_polynomial, 4)
regressor.fit(X, y, visualise=True)

bayes_regressor = BayesianLinearRegressor(0.1, noise_var, scalar_polynomial, 4)
bayes_regressor.compute_posterior(X, y, visualise=True)

log_res = LogisticRegressor(gaussian_rbf, X, 0.2)
log_res.fit(X_train, y_train, 0.005, 100, 64, X_val, y_val, visualise_training=True)

nn = NeuralNetwork(X.shape[1], [8, 8], ['sigmoid', 'sigmoid'], 'sigmoid')
nn.fit(X_train, y_train, 0.005, 8000, 16, X_val, y_val, 16, 'binary_crossentropy', visualise_training=True, save_name=None, display_metrics=True, compute_accuracy=True)

nn = NeuralNetwork(X.shape[1], [8, 8], ['relu', 'relu'], 'sigmoid')
nn.fit(X_train, y_train, 0.005, 8000, 16, X_val, y_val, 16, 'binary_crossentropy', visualise_training=True, save_name=None, display_metrics=True, compute_accuracy=True)