Bluetooth incompatible browser!

Do It Yourself Machine Learning

Welcome to the ultra:bit data trainer. Experiment and play with machine larning and build your first machine learning models - start here!

25 LED lights

"Record"-button

The website utilizes the bluetooth, accelerometer, LEDs, buttons, and sounds from the Micro:bit. Subsequetnly, you can connect the micro:bit to your computer via bluetooth without using the USB cable.

Try the expanded version

We have created an expanded version that, among other things, lets you modify the data representation and visualizes a k-nearest neighbor model.

You can finde the expanded version here: math.ml-machine.org

How does it work?

ML-Machine builds machine learning models to recognize patterns in the accelerometer data.

1. 

   With ML-Machine, you curate data samples of different gestures or movements, which a machine learning algorithm analyzes. It finds patterns in the data that can be used to predict/guess if the micro:bit is moved in circles, shaken, taped to the leg of the person jumping, etc.

2. 

   To achieve this, ml-machine.org makes a data representation of the samples by calculating specific properties of each sample, such as standard deviation, cumulated acceleration, and maximum value.

3. 

   The data representations of the data samples are shown to a neural network. The neural network “learns” from these examples by finding patterns in and between the samples.

4. 

   This training process results in a trained neural network (we also call it a machine learning model) that can guess/predict how the micro:bit is moved when the same properties (standard deviation, cumulated acceleration, maximum value, etc.) are calculated from the live accelerometer data.