urgent question,please about jule.torch HOT 8 OPEN

Hi, @zuoeye , did you specify the number of clusters at the beginning?

from jule.torch.

How to specify the numer of clusters at the beginning???

Thank you!

from jule.torch.

please see line 163 in train.lua

from jule.torch.

target_nclusters_table[i] = #(label_gt_table_table[i]) ?

Is this function‘s goal to get the clusters number of clustering?? This is not to specify the numer of

clusters at the beginning.

Thank you

from jule.torch.

Does I specify the number of target clusters ？ For example，should I specify four target clusters at the begining if I want to get four categories？

Thank you！

from jule.torch.

yes, you should. change #(label_gt_table_table[i]) to 4 in your case. Also take care of label_gt_tensor_table and label_gt_table_table at line 161 and 162, respectively.

from jule.torch.

Is the cvt2TabelLabels() function’s goal to get the numer of target clusters ？Why do we specify the

numer of target clusters if the function's goal is to get the numer of target clusters? I still don't

understand how to specify the number of target clusters? Could you help me with a detailed

description of how to modify the code to specify the number of target clusters？

Thank you！

from jule.torch.

I just made "target_nclusters_table[i]=4" of line 163. My results were very bad after I run this code in my data. I have changed the 'K_s' from 10 to 100, 'K_c' from 2 to 10 with 1, 'eta' from 0.1 to 0.9, and 'learningRate'. But the results were still bad. Do you know how to improve my results?

from jule.torch.