Simple Machine Learning Algorithms

Classification

A set of synthetic datasets and classification methods, inside the "classification" directory.

Binary Classification

Inside the "binary" directory:

• linear_batch.py contains a simple gradient descent algorithm made from scratch which is used to classify a two cluster, linearly-separable dataset.
• linear_stochastic.py uses Keras to create a single-layer (input and output) perceptron model and the Adam optimiser to classify a two cluster, linearly-separable dataset.
• nonlinear_stochastic.py uses Keras to create a multilayer perceptron model and the Adam optimiser to classify a two cluster, nonlinearly-separable dataset - more specifically, one cluster is inside another.

Multiclass Classification

Inside the "multiclass" directory, nonlinear_stochastic.py uses Keras to create a single layer perceptron model and the Adam optimiser to classify a five cluster, nonlinearly-separable dataset.

Perceptron

A set of simple neural network models made from scratch, inside the "perceptron" directory.

Single Layer

Inside the "singlelayer" directory:

• perceptron.py creates and trains a single-layer perceptron.
• error_plot.py visualises the loss over training rounds for the above model.

Multi Layer

Inside the "multilayer" directory, multilayer.py creates and trains a multilayer perceptron.