What is NEXT?
NEXT is a machine learning system, developed at UW-Madison, that runs in the cloud and makes it easy to develop, evaluate, and apply active learning in the real-world. Read more at the NEXT website.
How is NEXT useful for psychologists?
Estimate embeddings from triplet judgments.
Estimate a rank ordering on the basis of preferences.
Find the “best” item from a list of candidate items and user judgments (for example, the funniest cartoon).
Ok, I’m sold. How do I set up and run a psychology experiment using NEXT?
Our lab has prepared the following tutorials for replicating past NEXT experiments. The best approach is to walk through one of these replications and modify them to suit your own needs.
- How to replicate a triplet experiment
- How to replicate a ranking experiment
- How to replicate a “best candidate” experiment
Apart from these tutorials, the majority of NEXT how-to documentation can be found on its GitHub wiki pages at: https://github.com/nextml/NEXT/wiki. Also, because NEXT was developed outside of the Knowledge and Concepts Lab, technical questions that are not addressed by our pages here or are not in the GitHub wiki should be directed to the NEXT development team.
Citing NEXT
When using data collected with NEXT, please cite as follows:
Jamieson, Kevin G., et al. “Next: A system for real-world development, evaluation, and application of active learning.” Advances in Neural Information Processing Systems. 2015.