FAST: Feature-Aware Student Knowledge Tracing
This is the homepage of FAST, an implementation of HMM with features. FAST enables unsupervised learning with features.
We introduced FAST for modeling student learning (González-Brenes, Huang and Brusilovsky, 2014). FAST is an alternative to BNT-SM, a toolkit that requires researchers to design a different Bayes Net for each feature set they want to prototype. We think FAST is easier to use, because it only requires a CSV with the features you want to use. Also, the FAST toolkit is up to 300x faster than BNT-SM, and up to 25% better than conventional Knowledge Tracing.
We presented FAST at the 7th International Conference on Educational Data Mining (EDM 2014) (see slides ), where it was selected as one the top 5 paper submissions. In a follow-up study, we compared FAST to the best EDM paper of that year, with favorable results. This suggests that FAST has similar performance than custom, single-purpose models, but requires much less engineering effort.
Running FAST
Quick Start
- Download the latest release here.
- Decompress the file. It includes sample data for getting you started quickly.
- Open a terminal and type:
java -jar fast-2.1.1-final.jar ++data/IRT_exp/FAST+IRT1.conf
Congratulations! You just trained a student model, with student and item features, using state of the art technology.
For more details about the toolkit, see the Wiki.
Technical Details
FAST learns parameters for each skill using a modified EM algorithm for HMMs parameterized by logistic regression models (see slides, Berg-Kirpatrick et al, 2010).
Contact us
We would love to hear your feedback. Please email us!