TinyML

Chapter 4. The “Hello World” of TinyML: Building and Training a Model

In Chapter 3, we learned the basic concepts of machine learning and the general workflow that machine learning projects follow. In this chapter and the next, we’ll start putting our knowledge into practice. We’re going to build and train a model from scratch and then integrate it into a simple microcontroller program.

In the process, you’ll get your hands dirty with some powerful developer tools that are used every day by cutting-edge machine learning practitioners. You’ll also learn how to integrate a machine learning model into a C++ program and deploy it to a microcontroller to control current flowing in a circuit. This might be your first taste of mixing hardware and ML, and it should be fun!

You can test the code that we write in these chapters on your Mac, Linux, or Windows machine, but for the full experience, you’ll need one of the embedded devices mentioned in “What Hardware Do You Need?”:

• Arduino Nano 33 BLE Sense

• SparkFun Edge

• ST Microelectronics STM32F746G Discovery kit

To create our machine learning model, we’ll use Python, TensorFlow, and Google’s Colaboratory, which is a cloud-based interactive notebook for experimenting with Python code. These are some of the most important tools for real-world machine learning engineers, and they’re all free to use.

Note

Wondering about the title of this chapter? It’s a tradition in programming that new technologies are introduced with example code that ...