Additional material
On this page, you will find links and further material related to the topics on these pages.
Software
The online machine learning tool integrated into these learning units is an easy tool for creating neural networks, training them with your own data and then testing their performance.
The MemBrain software (free for non-commercial use; for Windows only) is a modelling tool that can be used to design, train and test neural networks.
With Teachable Machine from Google, networks can be created and exported as a TensorFlow.js model, e.g. for use in Scratch. Project examples for working with the machine learning extensions in Scratch and the Scratch-based programming platform PictoBlox can be found in our teaching materials “EcoKids Teach AI” (for primary school) and “STEM Resources 4.0” (for secondary school).
An enactive model
The material linked here (PDF) can be used to simulate a neural network as an enactive model with humans. The neural network consists of three layers and is suitable for recognising numbers. With the help of the instructions, 21 people and the experimenter take on the role of neurons in a human neural network. The neural network used here has already been trained and is able to recognise the numbers 1 to 8 on a 2x4 pixel grid. There is also a variant in which only 17 people + experimenter are required.
This activity is a highly simplified reduction of the “Stilwell Brain” simulated by Michael Stevens and Chris Eliasmith with over 200 people for their YouTube video (Mind Field Season 3 Episode 3).
Share this page
