These are chat archives for ramin-git/word_tree_structure

26th
Apr 2015
ramin-git
@ramin-git
Apr 26 2015 03:44
Good :). I understood that in your case you only devided words into bins which has 10 childs. For it your word_tree.th file structure like this
parent_id IntTensor(child_ids) // size = 10. These question about your case.
  1. Default value of Intensor what? Because may be bins can not be full or size of IntTensor not cerrectly 10, may be size is dynamic
  2. in your case root_id also has IntTensor(size=10) but if you only goes 1 iteration root_id must has Inttensor(size = 793471/10). İt appears that you do iteration until each parent_id has max 10 childs and last in iteration generating parent_id is root_id
    I could describe what i think?
Nicholas Léonard
@nicholas-leonard
Apr 26 2015 18:50
The size of bins can vary from bin to bin in a tree.
The leaf nodes (last ones down), are those without any children. That is how SoftMaxTree knows. The sizes don't really matter, as long as they aren't too big
ramin-git
@ramin-git
Apr 26 2015 19:22

ok, in my case (in example) tree will shown like this:

13 {11, 6} //90
14 {12, 3} //80
15 {8, 7} //70
16 {10, 2} //60
17 {1, 9} //50
18 {5, 4} //40
19 {13, 15, 17} //200
20 {14, 16, 18} //180
21 {19, 20} //380 (21 - it is root_id)