Fork me on GitHub


Helps you find a suitable neural network configuration for deep learning on time series.

470 commits | Last update: August 20, 2019

Cite this software

Choose a version:
[[ releases.length > 0 ? releases[selectedIndex].doi : conceptDOI ]]
Copy to clipboard
Choose a citation style:
Download file

What mcfly can do for you

  • Provides starting point for researchers to use deep learning
  • Creates deep learning models to classify time series data
  • Derives features automatically from raw data
  • Helps with finding a suitable model architecture and hyperparameters
  • Has a tutorial in Python to get you started!

Deep learning is a powerful tool to help with the automated classification of data. However, designing a deep learning network that works well for your data is not trivial: it requires the user to choose the number of layers in the network, the number of nodes in each layer, the type of each layer, and so forth. With so many degrees of freedom, finding the network that is right for your data is an arduous task. Moreover, each network still needs to be calibrated or trained before it can be usefully applied for automated classification of data.

mcfly simplifies this process by making explicit the required steps while offering useful default values at each step. mcfly then proceeds by trying out many different network configurations, training each one to the data provided by the user. It subsequently lists the performance of each network, along with a visualization that helps the user judge each network's tendency to overfit or underfit the data.

Read more
  • Machine learning
Programming Language
  • Python
  • Apache-2.0
Source code

Participating organizations


  • Christiaan Meijer
    Netherlands eScience Center
  • Dafne van Kuppevelt
    Netherlands eScience Center
  • Vincent van Hees
    Netherlands eScience Center
  • Patrick Bos
    Netherlands eScience Center
  • Mateusz Kuzak
    Netherlands eScience Center
  • Atze van der Ploeg
    Netherlands eScience Center
Contact person
Christiaan Meijer
Netherlands eScience Center