This project is a simple implementation of a neural network capable of learning the XOR logical function as well as recognizing digits in images.
The neural network is implemented in C++ and initially designed to learn the XOR logical function. It includes functionality for recognizing digits in images using the MNIST dataset.
The goal is to demonstrate the fundamental operations of a neural network and its ability to solve both classification and image recognition problems.
The neural network's architecture is customizable in terms of depth, width, and other parameters. You can adjust these settings in the AI.h file to customize the network according to your requirements.
To execute the program, follow these steps:
$ make
$ ./main XOR // Trains the network for XOR learning
$ ./main UXOR // Retrains the network for XOR learning
$ ./main MNIST // Retrains the network from scratch for image recognition using the MNIST dataset
$ ./main UMNIST // Retrains the network for image recognition using the MNIST dataset
$ ./main TXOR // Tests the trained neural network with XOR values
$ ./main TMNIST1 // Tests the trained neural network with the MNIST training dataset
$ ./main TMNIST2 // Tests the trained neural network with a custom image located in the img/test_img.png directory
$ make clean
You can directly test the network after training by appending the testing command. For example:
$ ./main UMNIST TMNIST2
The network will be retrained for XOR learning, and then it will be tested with a custom image.
The weight files generated during training are stored in the train/ directory.
This project is licensed under the MIT License. See the LICENSE file for more details.