No labels - When stop training

michaal94 / torch_DCEC

Pytorch Deep Clustering with Convolutional Autoencoders implementation

MIT License

98 stars 30 forks source link

No labels - When stop training #10

Open fernandorovai opened 3 years ago

fernandorovai commented 3 years ago

Thanks for sharing this - I'm training on a non-labeled dataset, how do I know that I should stop training if I cannot calculate the metrics? Also, my Total Loss is decreasing, but still bigger than one, is it a problem? Thanks!

kroegern1 commented 1 year ago

Agreed, why is labels assumed to be in the dataset? They're eventually needed for updating the loss, I don't think this is the case in the original method