Diversity Criteria

Two Birds, One Neural Network
1

Hi
Can you please explain how is the diversity criteria scored,

2

Hello @balram0102

The main idea behind diversity metrics is to ensure that your X0/X1 lines comprehensively cover the potential area with optimal density.
I hope it is helpful.

Kind regards,
Onward Team

3

@discourse-admin May I ask what is the maximum score? Is it 1.0?

4

@discourse-admin

The main idea behind diversity metrics is to ensure that your X0/X1 lines comprehensively cover the potential area with optimal density.

I have another question. Does it means that the metric is computed in the x space relative to the hidden x_true, rather than in the reconstructed y space in comparison to the given y_true?

1 Like
5

Hello @daisuke0530
Yes, you’re totally right, the maximum score is 1.0.

Best regards,
Onward Team

1 Like
6

Hello @daisuke0530,
Indeed, the diversity metric is specifically designed to assess the coverage of the X space by the generated x.
Best regards,
Onward Team

1 Like
7

I think the X space is bounded between 0 and 1 (atleast that’s what it looks like from the data). If so, if the submission has samples whose x values are out of X space boundaries, does it mean the sample is not following the criteria?

Regards,
Vishwas

8

Hello @chepurivishwas360

Yes, you’re correct about the boundaries for the X space. Additionally, you should be aware, that If the x values are outside these boundaries, they are considered as bounded - with 0 for all x values less than 0, and 1 for all x values greater than 1.

Kind regards,
Onward Team

1 Like