Multilayer Perceptron

🔤🧠 Implementation of a simple artificial neural network in the form of a perceptron, which can be trained on an open dataset and perform recognition of 26 handwritten letters of the Latin alphabet.

Install:

1. Go to /src
2. Do make install
3. Run from Desktop icon (or run executable ${PWD}/multilayer_perceptron/MultilayerPerceptron.app)

Overview:

• The program supports two types of internal implementation: matrix and graph. This can be configured in the main menu of the program.

• If you change the number of internal layers or the implementation, you should load the appropriate weights or train the neural network again. The Weights folder contains ready-made weights for two-layer perceptron.

• Perceptron uses the EMNIST LETTERS dataset for training.