sksom / Classification-using-KMM-Kernel-Mean-Matching-

Dataset Bias correction (Python)
19 stars 1 forks source link

urgent problem #1

Open zhangjiantianyasmile opened 5 years ago

zhangjiantianyasmile commented 5 years ago

I am interested in KMM, and I downloaded your code and ran in my computer. Could you explain one doubt about your code. Thank you very much! My doubt as follow: In buildModel() function, we can get a tuple (acc, acc2) where acc is get from sample_weight=beta and acc2 is get from sample_weight=beta_fixed. But in the end of main() function , print Accuracy without KMM is res1(acc) while print Accuracy with KMM is res2(acc2). May I have wrong understanding or other something? I am sorry to bother you and look forward to your reply. Thank you very much.

stephenxieber commented 2 years ago

Actually, res1(87.5%) is the result with KMM, while res2(93.75%) is the result without KMM. There are some problems in your code