No description, website, or topics provided.
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
LICENSE
README
active.py
crowd_model.py
do_test.py
gzoo.py
plot.py
results_correlated.py
run.py
start.py
util.py
util2.py

README

Source code for our paper:

An T. Nguyen, Byron C. Wallace, and Matthew Lease. A Correlated Worker Model for Grouped, Imbalanced and Multitask Data. In Proceedings of the 32nd International Conference on Uncertainty in Artificial Intelligence (UAI), 2016. 

We are not allowed to release the datasets (RCT and Galaxy Zoo 2) but they should be available on request from their owners.

To run on your data, you will need a crowd_model.label_collection object and pass it to 
 - crowd_model.tc_model  : to run Two Coin Model (Raykar et. al. 2010)
 - crowd_model.mv_model  : to run Majority Voting
 - crowd_model.hc_model  : to run Hybrid Confusion Model (Liu & Wang 2012) (this is the same model as Two Coin except for inference by MCMC, which we implemented by PyMC)
 - crowd_model.vss_model : to run the correlated sensitivity-specificity model (with variational inference) presented in the paper


To run the multitask model: you will need crowd_model.multitask, which is initialized by two vss_model (one for the source task, one for the target task)

See crowd_model.py for more details.