search

slaypni / fastdtw

A Python implementation of FastDTW
MIT License
777 stars 121 forks source link

Local step size condition #9

Open ronggong opened 7 years ago

ronggong commented 7 years ago

Hi slaypni,

Thanks for your great work! In my work, I am experimenting with different local step size, such as (2,1)(1,2)(1,1). Could you point out where to change these step sizes in your cython code?

slaypni commented 7 years ago

@ronggong Sorry for this late response. I do not understand what the local step size is. If it can be added to FastDTW algorithm, it may be better to start by modifying fastdtw.py. fastdtw.py is simpler than cython code, and similar to the pseudo code written in FastDTW paper.

ronggong commented 7 years ago

Thanks for the response. I somehow need a fast calculation speech, the python version is too slow for my implementation which will calculate a bunch of pair-wise dtw similarities. I failed to get your cython version fastdtw running. Maybe it was the compilation problem, the default fastDTW implementation is python version.

slaypni commented 7 years ago

As you mentioned, there is an issue for installing cython version of fastdtw at this moment. There might be some problems in setup.py, however, I did not figure it out yet.