perceptron-from-scratch

Designing a simple discrete perceptron algorithm from scratch, without the use of external ML libraries.
(current version: 0.1)

Author: Siddharth Yadav (syntax-surgeon)

main.ipynb

• Contains version-0.1 of the main program which tries to implement a discrete perceptron algorithm
• The core of the algorithm is based on an intuitive implementation of basic gradient descent
• The current version of the codebase is functional, however, future versions may be OOP-based
• Simple plotting functionality is included to visualize the results of the training

Figure showing the dataset with the trained perceptron (left) and the mean absolute error per epoch (right)

Features planned for release 1.0

• Converting the codebase to an OOP architecture
• Inclusion of additional perceptron training features such as 'strict_reduce' and 'descent_type'
• Advancing the graphing/plotting functionality to include intermediate perceptrons
• Providing a choice a datasets for testing purposes
• Increasing the verbosity of the in-code documentation