Using C++ I Implemented Multilayer Perceptron. Using this project you can create a custom model with multiple layers, different activation functions, with a specified input shape based on your own data. Set the learning rate for your model and fit the data to the network, all the feed forward and back propagation is executed by the class. After training your network, you can evaluate the test data and check the network performance.
You can also save the network and load it later in a different program.
- Matrix class for defining each layer's weights and their matrix calculations methods
- Layer class for storing weight values with different neuron activation methods
- Neural Network save and load method
You can look up Visual Studio and OpenCV setup from here.
Here is one sample for working with the Neural Network class.
# | Sample Name | Description |
---|---|---|
1 | Multiclass Classification | A one layer neural network for classifying five groups of data with two values |