top of page
Artificial Neural Network From Scratch
-
Building Artificial Neural Network (ANN) from scratch using C#.net.
-
Developing tools with user interfaces to understand single perceptron and back-propagation.
-
Adapting the ANN to recognize simple characters and faces.
Inspiration from Neurobiology
Face Recognition
The ANN can be trained on 8 different faces taken by the webcam. Then in the testing the error of the input image being in each of the 8 trained faces are displayed!
Character Recognition
Any characters of 8x10 pixels can be constructed through mouse click and used as input for training and inference!
Visualization of Weights and Backpropagation
The tool is used to visualize the training process of a ANN model of 4 input and 2 output layers. The weights and activations at each layers are visualized.
Single Perceptron
visualizes a single node (perceptron) which can be trained to work as simple gates such as OR, AND etc.
bottom of page