Is the "test_accuracy" in test_metrics.csv in output files already the mean accuracy of Leave-One-Subject-Out CV?

wenhui0206 / NeuroGPT

Code for the paper "Neuro-GPT: Towards a Foundation Model for EEG"

GNU General Public License v3.0

110 stars 22 forks source link

Is the "test_accuracy" in test_metrics.csv in output files already the mean accuracy of Leave-One-Subject-Out CV? #7

Open chocooni opened 4 months ago

chocooni commented 4 months ago

Hi,

As in subject, I'm curious about is the "test_accuracy" in test_metrics.csv in output files already the mean accuracy out of 9 subjects using Leave-One-Subject-Out CV?

Thank you.

Also I think in train_gpt.py, line 155:

test_dataset = train_dataset

seems confusing, is it a mistake?

henrisuurorg commented 2 weeks ago

test_dataset = train_dataset

@chocooni Did you ever find out if this was intentional? It's confusing for me as well.

Any comment @wenhui0206 ?

wenhui0206 commented 2 weeks ago

Hi, sorry for the confusion. You can ignore the test_dataset here. The real testing phase (fine-tuning) is a separate run which is defined in scripts/finetune.sh.