Simple Networks

program

Submitted by

Credits

IMAGINARY

Contributors

Supported by

Carl Zeiss Foundation

An exhibit to visualize and playfully discover the mechanisms of neural networks.

Using the example of minimalist neural networks, “Simple Networks“ playfully illustrates the mechanisms that take place inside an AI. Through various levels you can explore these mechanisms intuitively and train an AI step by step yourself. Visual cues will help you achieve good training results. You can change the input values and weights simply by dragging them. At the beginning you will only be manipulting a few neurons, but with each level the tasks become more difficult and complex. You will see hidden neurons and bias, both of which are also adjustable. The aim of each level is to hit the desired output value (for all training data) as accurately as possible.

Main File

Repository

Files