After searching quite some time for it on Google I could not find a sufficient software/toolbox that can manage trainings of neural networks. I thought of a program that combines visualization techniques without the need to write code as well as having the possibility to compare several trainings of neural networks and be able to store them easily. Does a program like this exist?

Regards Lukas


It does, it's an API called Keras, written on top of TensorFlow. With that you can rapidly prototype any "standard" NN architecture, such as feedforward NN, RNN, or CNN. With it you can export the trained model with one line of code and save it for later use. As for visualization, Keras offers integration with pydot (GraphViz) to plot the whole NN as a graph for you to inspect.

  • $\begingroup$ I have used Keras for a short time, but I thought more of a GUI application that visualizes and compares different networks. Is that possible in Keras? $\endgroup$ – Lukas Jun 4 '18 at 20:57
  • $\begingroup$ Define "compare". If you mean visual comparison of model architecture, you can export the two models as png graphs and visually inspect them. Source: keras.io/visualization $\endgroup$ – pcko1 Jun 4 '18 at 20:58
  • $\begingroup$ I was talking about comparing different metrics of model trainings and hyperparameters used for the training. $\endgroup$ – Lukas Jun 4 '18 at 21:42

After further searching online I found these useful packages:

  • sumatra
  • sacred
  • FGLab
  • CDE

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.