Artificial Neural Network for Arduino called Neurona, tested on Arduino UNO, and MEGA. Neurona is an Arduino library which allows boards to feed Artificial Neural Network (ANN) structures in order to perform tasks such as pattern recognition (classification), non-linear regression, function approximation and time-series prediction from the Multi-Layer Perceptron (MLP) architecture implementation.
In this example, it is a 2-bit binary classification using Neurona library. It is easy and simple to do.
The following steps to implement Neural Networks on Arduino are
-
Download the Neurona library and add it to your Arduino IDE.
-
Go to MLP Topology Workbench.
-
Click TRAINING tab and input the dataset like this
-
Click TOPOLOGY tab and make your neural netowrks of 2 inputs, 1 hidden layer with 4 neurons, and 4 outputs.
-
Train your Neural Network model. Go back to TOPOLOGY, click update parameters then click iterate. Wait for a while, after training process, put some data at input nodes and see what will happen to output nodes.
-
Then click save, it will show the weights for each neurons. You will use this to Arduino sketch.
-
Prepaper your Arduino sketch. I already did this for your, just hit upload! Then input the data (00, 01, 10, 11) in the Arduino IDE serial monitor.
It so easy! You can implement it to complex data for classification for an example color recognition.