search

mlgig / mrsqm

GNU General Public License v3.0
28 stars 8 forks source link

Best Parameters on UCR-Datasets #6

Closed patrickzib closed 1 year ago

patrickzib commented 1 year ago

Hi, great work! :)

I am having problems to set the best hyper-parameters for the UCR-datasets.

Currently, the default-parameters seem to be optimized for speed, but the accuracy is rather low (measured on my hardware)

Total mean-accuracy: 0.814 Total fit_time: 3257.89 Total pred_time: 1965.34

What would be the best set of parameters to obtain get results from the paper? In the paper this is referred to as MrSQM_SFA_k5.

lnthach commented 1 year ago

Hi Patrick,

Thank you for your interest. In the paper we set nsax = 0 and nsfa = 5 and default values for the remaining parameters. We find this the most balanced configuration in our experiments: not slow and good accuracy. Increasing nsfa (>5) can improve the overall accuracy a little but it is not worth the cost of speed in our opinion.

patrickzib commented 1 year ago

Hi Thach,

great, thank you :) I was not sure how to translate SFA_k5 into nsax and nsfa. Your answer solved it.