This is an old side project I did in my first year of learning programming, when I was considering data science as a career.

Python implementation of common machine learning algorithms. Currently supports neural networks, logistic regression and linear regression. Datasets are expected to be Numpy arrays (Pandas dataframes count).

Offers train_network to build a neural network on training data. Once trained, function returns a NeuralNetModel object with a get_predictions method for generating predictions on new datasets. "Neural Network.ipynb" contains an example of a network trained on the MNIST dataset of handwritten digits.

Offers gradient_descent to train a linear model. Returns a vector of weights that can be used with the get_predictions method to generate predictions on new data.

Offers two implementations of logistic regression, BFGS and gradient_descent, using those respective methods. Each method returns a vector of weights to be used with the get_predictions method.

Offers two methods get_binary_data and get_dense_data. Each method creates a dataset with normally distributed features, with either a linearly separable vector of labels (for binary data) or a linear combination of the features (for dense data) with randomly chosen weights. Weights are returned alongside the data for comparison with weights trained by a model.