search

baharanm / craig

Data-efficient Training of Machine Learning Models
60 stars 24 forks source link

How to find a coreset for a certain epsilon (training residual error)? #4

Open vanshajkhattar opened 1 year ago

vanshajkhattar commented 1 year ago

Hi @baharanm,

I really enjoyed reading your paper, and thank you for making the code publically available. I was wondering if it is possible to find a minimum size coreset for some given \epsilon (in eq. 12 in your paper) using your code?

Thank you!

vanshajkhattar commented 1 year ago

Hi @baharanm,

I also have one follow-up question, and would highly appreciate it if you could provide your input. Is there a way that I can modify the given code for regression settings, where I do not have the number of classes defined? It looks like the algorithm needs knowledge of the number of classes to compute a coreset.

Thank you for your time! Vanshaj

ndvbd commented 8 months ago

@vanshajkhattar can you see where in the paper the algorithm for neural network is described clearly? In alg 1, the F and L are not defined, You have equations 8 and 16, but they are for some w. Is there a clear applicable algorithm somewhere?

Abhicoder03 commented 4 months ago

@vanshajkhattar Hey, Just read this paper, and I want to implement it but i find codes a little bit confusing. Could you please help me in this. It would be great.