stocknet-code icon indicating copy to clipboard operation
stocknet-code copied to clipboard

Published 20 hours ago • verified publisher icon yumoxu

cant get 0.58 acc

Open nickhuangxinyu opened this issue 6 years ago • 14 comments

Hi, yumoxu.

great work about your paper, but i tried to reproduce the results you mentioned in your paper.

only 0.53 i can get.

Is there any tricks i should know to get 0.58?

Thanks

nickhuangxinyu avatar Dec 09 '18 08:12 nickhuangxinyu

I have the same question !

thtang avatar Jan 14 '19 06:01 thtang

I got acc: 0.574405 without doing anything special. Not 0.58 acc still, but I guess we can get 0.58 with more trials?

Chih-Ling-Hsu avatar Apr 17 '19 08:04 Chih-Ling-Hsu

@Chih-Ling-Hsu I wonder what things you did. I run the code with GTX 1080Ti and default config,yml, only get 0.435516 in the test set.

crazyofapple avatar Apr 18 '19 12:04 crazyofapple

I also run the code with GTX 1080Ti and default config.yml, following the instructions in README file. I have tried a second time and still get 0.57 acc so I guess it was not just a rare case.

Chih-Ling-Hsu avatar Apr 19 '19 02:04 Chih-Ling-Hsu

@Chih-Ling-Hsu
Have you checked the majority of the stock movement of the pre-processed data?

p-null avatar May 19 '19 07:05 p-null

不管怎么调，结果每次都是 0.5744。有问题！

hjc5858 avatar Jan 15 '20 16:01 hjc5858

@Chih-Ling-Hsu I wonder what things you did. I run the code with GTX 1080Ti and default config,yml, only get 0.435516 in the test set.

Mine is 0.483135 not much different both in py27 and py36 with CUDA10

bahung avatar Jun 26 '20 11:06 bahung

I also run the code with GTX 1080Ti and default config.yml, following the instructions in README file. I have tried a second time and still get 0.57 acc so I guess it was not just a rare case.

I tried with Tesla P100 and got the same result with you by 30 epochs.

bahung avatar Jun 30 '20 01:06 bahung

i have the same question

zqy1106 avatar Aug 28 '20 09:08 zqy1106

@Chih-Ling-Hsu I wonder what things you did. I run the code with GTX 1080Ti and default config,yml, only get 0.435516 in the test set.

Mine is 0.483135 not much different both in py27 and py36 with CUDA10

Hi, I'm running this repo with CUDA10. It takes a long time to finish one epoch. Did this happen in your test?

WJMacro avatar Nov 24 '20 08:11 WJMacro

I got acc: 0.574405 without doing anything special. Not 0.58 acc still, but I guess we can get 0.58 with more trials?

The test dataset has 1008 samples, 579 of which are positive. 0.574405=579/1008. Maybe you should check your confusion matrix to see whether all samples in the testset were predicted to be positive.

WJMacro avatar Nov 30 '20 07:11 WJMacro

does anyone here have this implementation on pytorch or tensorflow2 ? i'm a beginner i want to run this on colab but it doesn't work since it required tensorflow 2

sia-watsonlee avatar Sep 15 '22 20:09 sia-watsonlee

不管怎么调，结果每次都是 0.5744。有问题！

请问vocab.txt是什么呢，怎么下载呀

Sia2010 avatar Oct 01 '23 04:10 Sia2010

Hi, yumoxu.

great work about your paper, but i tried to reproduce the results you mentioned in your paper.

only 0.53 i can get.

Is there any tricks i should know to get 0.58?

Thanks

Hi, could you plz tell me what the vocab.txt refer to? Where can I download it?

Sia2010 avatar Oct 11 '23 13:10 Sia2010