This project implements a handwritten digit recognition system using RSM (Reservoir State Machine) propagation.
The system focuses on simplicity and efficiency, making it a useful experimental and educational approach compared to more conventional or advanced deep learning methods.
- Recognizes handwritten digits (0–9)
- Uses reservoir-based propagation instead of backprop-heavy deep networks
- Lightweight and efficient for experimentation
- Modular design for easy extension and testing
- Works with a dataset of handwritten number images
- Normalized (grey-scaled, resized)
- Flattened (converted into a single sequence representing the image)
- All nodes (except for the input layer) are created with random bias and weight, value and error margin are set to 0 (calculated with each update)
- Nodes are organized into layers, and layers into the neural network object
- The input is propagated through the neural network using iterative state updates
- Each input influences the neural network's state
- Previous states contribute to future states
- This creates a state that encodes features of the handwritten digit
- Based on the previous layers' data, nodes in the output layer are given values
- The output node with the highest value is the most probable choice according to the network
- The actual value of the image is compared to the determined output, and the margin of error is propagated backwards through the network, and all values are updated accordingly
Results of machine learning algorithm generated by Chat-GPT
Results of RSM Back-propagation Algorithm On 100 Images
Results of RSM Back-propagation Algorithm On 1000 Images