Is there a tool to visualise neural networks similar to how TensorFlow playground does it? It is a custom tool for beginners, but I was wondering if there's a similar tool. It shows you what effect each neuron has on the model, the weights of each neuron, and you can see the information passing through in a feed-forward neural network.
You can create wanted visualisation by yourself on python. For example, this is my jupyter notebook with spiral classification problem like at the playground: https://github.com/Belerafon/Spiral-2.0/blob/master/Spiral.ipynb .At the final step it creates picture of network activation like at the playground.
Visualisation capabilities shown at the playground limited by two-input (2D) case of task. In all real cases with multi-dimensional input data (like pictures) there are no simple way to show neuron activation pattern. It is possible only show active neurons for current input and you can try to find interesting for you neurons by changing input samples. For example this researchers created Deep Visualization Toolbox for convolutional network: https://www.youtube.com/watch?v=AgkfIQ4IGaM