In The Shed Issue No. 105, we saw how machine learning (ML) could be embedded in a smart device such as the Arduino Nicla board. The following steps can be taken to activate the prediction features in our board:
1. Acquire data through sensors — usually part of the ML board creating a collection of samples.
2. Use the samples to define the data model with the ML platform — like Neuton.AI in our case.
3. Test the model with sample data on the platform using the Web.
4. Generate the software to embed the prediction features in the ML microcontroller.
5. Run the prediction offline on the programmed ML microcontroller.
This sounds very impressive, but a final problem arises: how to manage the microcontroller sensors to