Course:CPSC532:StaRAI:2017:Matt
Results for Predicting Gender from Movie Ratings
MLNs with Anglican:
| Method | ml-60k ASE | ml-60k Log loss | ml-1m ASE | ml-1m Log Loss | Yelp ASE | Yelp Log Loss |
|---|---|---|---|---|---|---|
| Predict 0.5 | 0.25 | 1 | 0.25 | 1 | 0.25 | 1 |
| training average | 0.2159 | 0.9004 | 0.2043 | 0.8637 | 0.2364 | 0.9604 |
| Anglican, r>=4, no hidden units, weights per item | 0.2333 | 0.9515 | 0.2241 | 1.0891 | 0.2461 | 0.9886 |
Here, the Anglican results use SMC inference (with 5 particles) for ml-60k, and IPMCMC for ml-1m and Yelp. Both inference algorithms are described in:
@ARTICLE{Anglican2015,
author = {{Wood}, Frank and {van de Meent}, Jan Willem and {Mansinghka}, Vikash},
title = Template:A New Approach to Probabilistic Programming Inference,
journal = {ArXiv e-prints},
archiveprefix = {arXiv},
eprint = {1507.00996},
primaryclass = {stat.ML},
keywords = {Statistics - Machine Learning, Computer Science - Artificial Intelligence, Computer Science - Programming Languages},
year = {2015},
month = {jul}
}
Anglican reference is here:
@article{Anglican2016design,
title = {Design and Implementation of Probabilistic Programming Language Anglican},
author = {Tolpin, David and van de Meent, Jan Willem and Yang, Hongseok and Wood, Frank},
journal = {arXiv preprint arXiv:1608.05263},
year = {2016}
}