evelinehong / lbf Goto Github PK

Hi, I read your paper and it was very impressive. Thanks for publishing cool research and implementation.
The code seems working well, but I'm struggling in a small problem.
Can I get trees before embedded? I checked several variables and most of them was embedded information which is non-interpretable. I want visualized trees or readable mathematical expressions in training or testing steps.
It will be glad to receive an answer :)

I ran the model (100 epoch, 50-fix, ma-fix) and got 53.2%. According to the paper, the setting the same as mine is 59.4% (the heighest accracy of the test is 100-fix, 59.6%). Does additive condition existed for 59.4%? I just 53.2%
The setting in your code is

I did not change.

Does the models (models/generator, merge,....) you provided is well-training？Can I use it to test directly?

Hi,
I ran the training code using "python run_seq2tree.py --model ma-fix --nstep 50 --name ma-fix". But I'm getting best test accuracy of around 53.6%. Can you please provide the exact details of the setting in which 59.4% accuracy was obtained?

Could you please provide the training code which initializes the memory buffer with all the ground-truth expressions? I have tried to re-produce, but I cannot reach 74.1 which is reported in your paper.

What is the meaning of the "ma-fix" model? What is the difference between fix and ma-fix?
就是那个ma-fix和fix有什么区别啊靓女