How are the hidden target set normalized?

Hi @discourse-admin,

I would like to know how the hidden targets for the test set are normalized and how the score is calculated, is it a simple minmax normalization to scale them in 0-255, or can they have a min/max of other than 0/255?

Thank You

Hi @harshitsheoranlearnt

The test and hold out dataset are indeed scaled from 0-255 using a min-max scaler.

ThinkOnward Team

1 Like

Hi @discourse-admin,

Thank you for your reply, Is the normalization on the test and hold out set done on a slice level or volume level, in other words, was it done with sample code:

  1. minmax_normalize(array[:, 75])
  2. minmax_normalize(array)[:, 75]

Was it done 1. or 2.?

Hi @harshitsheoranlearnt

The min/max values for rescaling (0-255) are obtained on volume level and not a slice level.

ThinkOnward Team

2 Likes

Thank you, That’s what I was missing earlier, I confirmed it with a sub today :smiley: