akundaje

About

Username
akundaje
Joined
Visits
2
Last Active
Roles
Member
Points
2
Badges
1

Comments

  • Looks great. We use similar strategies with forcibly adding minority class samples to each minibatch and weighted loss. But still trying to figure out a loss function that explicitly optimizes precision/recall rather than auROC or cross entropy. Not…
  • Ok great. But does your test set reflect the actual imbalance you expect to see when you use this to call variants across the whole genome? If not, then the test set would not be representative.
  • What's your class imbalance for your classification tasks. I expect there is significant imbalance since variants are rare events as compared to the size of the genome. If that is the case, auROC can be very misleading. You'd want to report area und…