bisno / WDRO

UniDT鲁棒优化算法 The implementation on Towards Scalable and Fast Distributionally Robust Optimization for Data-Driven Deep Learning.
MIT License
181 stars 26 forks source link

multinomial classification #7

Open laohuu opened 1 year ago

laohuu commented 1 year ago

How we can divide training data for training multinomial classification?

bisno commented 1 year ago

When using Label-Distribution Partitioning, we have the samples of subgroup that follow label distribution. For 'Partitioning Evenly' method, each subgroup has same amount of samples.