search

ml-jku / DIffUCO

19 stars 3 forks source link

Confusion about unsupervised learning #2

Open chjchjchjchjchj opened 1 week ago

chjchjchjchjchj commented 1 week ago

What a nice job! But I have some questions about the concept of "unsupervised". From your codebase, I see that you use GurobiSolver to generate some solutions and energy as the labels, isn't it supervised learning?

sanokows commented 1 week ago

Thanks! The solutions from gurobi are only used to calculate the difference between the model solutions to the optimal energy value. This difference is then logged to wandb (e.g. Relative energy). These solutions are never used for model training. I hope this answer your question.

Best,

Sebastian Sanokowski

Haojun Chen @.***> schrieb am Mi., 26. Juni 2024, 17:26:

What a nice job! But I have some questions about the concept of "unsupervised". From your codebase, I see that you use GurobiSolver to generate some solutions and energy as the labels, isn't it supervised learning?

— Reply to this email directly, view it on GitHub https://github.com/ml-jku/DIffUCO/issues/2, or unsubscribe https://github.com/notifications/unsubscribe-auth/AEW4JJSD7DAU27UWFEAZNCLZJLMRPAVCNFSM6AAAAABJ6ELUYKVHI2DSMVQWIX3LMV43ASLTON2WKOZSGM3TKNRVGQ3TOMA . You are receiving this because you are subscribed to this thread.Message ID: @.***>