Make features and annotations into experiment data

Question

Make features and annotations into experiment data

jixu7952 opened this issue 7 years ago · comments

jixu7952 commented 7 years ago

missing script?

Jen-Yu Liu · Answer 1 · Thu Feb 23 2017 18:13:41 GMT+0800 (China Standard Time)

The features of all 6 scales from MagnaTagATune will take quite a lot of spaces. I will see what I can do.

jixu7952 · Answer 2 · Fri Feb 24 2017 08:54:48 GMT+0800 (China Standard Time)

a few sample lines of data to show the npy file format would be desirable, and the python file for making experiment data is missing

TobyOoO · Answer 3 · Fri Feb 24 2017 11:52:59 GMT+0800 (China Standard Time)

Yes, I am confused about how to retrieve tags backward from outcomes of the model.

Jen-Yu Liu · Answer 4 · Fri Feb 24 2017 12:43:35 GMT+0800 (China Standard Time)

Thank you both for your comments. I think I will take this chance to overhaul the whole package, including updating the codes, uploading the experiment data.

I will also make the feature standardization process independent of the scikit-learn package by including the standardization directly into the model. The standardizer module in scikit-learn has been changing several times recently, so doing so will make the maintenance easier in case it changes again.

jixu7952 · Answer 5 · Mon Feb 27 2017 13:31:15 GMT+0800 (China Standard Time)

thanks for update, Ciaua. Could you please kindly share the script to prepare the experiment data?

Jen-Yu Liu · Answer 6 · Mon Feb 27 2017 14:11:42 GMT+0800 (China Standard Time)

I have added process_annotations.top_n_tags.py and make_exp_data.py.
You still have to download the MagnaTagATune files from their website though.