This code offers a Python implementation of the work presented in:
Josif Grabocka, Nicolas Schilling, Martin Wistuba, Lars Schmidt-Thieme (2014): Learning Time-Series Shapelets. In Proceedings of the 20th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD 2014
This implementation builds upon the keras library (basically, you will need keras, tensorflow and numpy to be
installed) for efficient optimization of the Shapelet coefficients.
As an example, it takes roughly 1 minute (on a standard MacBook Pro laptop) for training on the Trace dataset from UCR/UEA repository.
This code is now integrated into the tslearn toolkit.
Have a look there if you are interested.