How to tackle uncertainty in pediatric ALL subtype classification? Our group combines RNA-seq data with machine learning to classify molecular ALL subtypes --- But uncertainty in predictions has always been a challenge... until now 🩸 1/4
Comments
Log in with your Bluesky account to leave a comment
This approach helps us to:
1) To quantify the uncertainty of individual classifier predictions;
2) To provide prediction sets that control the false negative rate; and
3) To reduce classifier errors, transforming incorrect predictions into uncertain predictions. ‼️
3/4
As part of these results, we developed and made available the Python package "conformist", which facilitates the application of CP to softmax scores from any classifier. Check out our preprint for details 4/4
Comments
1) To quantify the uncertainty of individual classifier predictions;
2) To provide prediction sets that control the false negative rate; and
3) To reduce classifier errors, transforming incorrect predictions into uncertain predictions. ‼️
3/4