Dear colleagues, We are pleased to announce the release of PRISM2.3 which is available for download at:
http://rjida.meijo-u.ac.jp/prism/ PRISM is a probabilistic extension of Prolog that has been developed as a generic interface between users and probabilistic modeling. Not only does it support generative modeling by offering a dazzling array of built-ins for Bayesian and non-Bayesian inference, it also provides built-ins for discriminative modeling such as CRFs and those for modeling cyclic relations that lead to infinite probability computation. The latest PRISM, PRISM2.3, further adds a unique mechanism of ranking-based parameter learning. Now the user can learn probabilities by giving the system a ranking of propositions and obtain parameters such that high-rank propositions have higher probabilities than low-rank propositions. This characteristic is applied to the situation where ranking is more important than probability itself such as user preference and information retrieval. We believe that PRISM2.3 now offers the widest flexibility of parameter learning for logic-based probabilistic modeling for Bayesian and non-Bayesian users. With Best Regards, Taisuke Sato, Yoshitaka Kameya, Ryosuke Kojima, Neng-Fa Zhou
_______________________________________________ uai mailing list uai@ENGR.ORST.EDU https://secure.engr.oregonstate.edu/mailman/listinfo/uai