Skip to content

estimate uncertainty using dropout and conformal prediction

Notifications You must be signed in to change notification settings

appeler/uncertain_names

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

57 Commits
 
 
 
 
 
 

Repository files navigation

uncertain names: uncertainty using dropout and conformal prediction

Using the FL voter registration data and the base LSTM model, we illustrate how to produce uncertainty estimates using dropout and conformal prediction.

Side-by-side

We randomly picked 10,000 unseen names (not in the training set) from the test set to illustrate how outputs differ across these cases. The output CSV has the following columns: last_name, label, conformal_pred_set, lw_ci_nh_white, up_ci_nh_white, ...

Here's a short snippet of the output:

Application

We compare the average CI and the average size of the prediction set (via conformal inference) of seen names to unseen names as one of the applications of uncertainty estimates. As the following notebook illustrates, the average set size and average CI is slightly smaller for seen names.

Authors

Bashar Naji and Gaurav Sood

About

estimate uncertainty using dropout and conformal prediction

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published