-
Notifications
You must be signed in to change notification settings - Fork 154
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
How to decide number of candidates #56
Comments
If I try to address this increasing candidate size will increase accuracy and will be less sensitive to anomalies vise versa. threshold here is window size. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
I trained the model on my custom dataset in which each line contains 3 events (e1, e2, e3) (a tri-gram) where I want to predict e3 based on e1 and e2. An example is shown below:
![image](https://user-images.githubusercontent.com/30209002/175777242-510c43cc-34e9-4957-9c31-12b800a094ca.png)
![image](https://user-images.githubusercontent.com/30209002/175777315-191fec8c-37a2-4a7b-aead-dec71c02cc99.png)
After training for 50 epochs I got:
On the predict stage, I kept
![image](https://user-images.githubusercontent.com/30209002/175777482-40f9a4ca-1349-465d-bee0-a82a76e10afe.png)
num_candidates=1
, then I got very low F1-scorenum_classes
in my dataset are 9. If I increase, num_candidates, divide by zero error occurred. What is the best approach to choose num_candidates for given dataset?Thanks
The text was updated successfully, but these errors were encountered: