eprintid: 1470010 rev_number: 51 eprint_status: archive userid: 608 dir: disk0/01/47/00/10 datestamp: 2018-05-08 13:14:39 lastmod: 2021-10-10 22:38:24 status_changed: 2018-05-08 13:14:39 type: proceedings_section metadata_visibility: show creators_name: Jitkrittum, W creators_name: Gretton, A creators_name: Heess, N creators_name: Eslami, SMA creators_name: Lakshminarayanan, B creators_name: Sejdinovic, D creators_name: Szabó, Z title: Kernel-based just-in-time learning for passing expectation propagation messages ispublished: pub divisions: UCL divisions: B02 divisions: C08 divisions: D76 divisions: B04 divisions: C05 divisions: F48 note: This version is the author accepted manuscript. For information on re-use, please refer to the publisher’s terms and conditions. abstract: We propose an efficient nonparametric strategy for learning a message operator in expectation propagation (EP), which takes as input the set of incoming messages to a factor node, and produces an outgoing message as output. This learned operator replaces the multivariate integral required in classical EP, which may not have an analytic expression. We use kernel-based regression, which is trained on a set of probability distributions representing the incoming messages, and the associated outgoing messages. The kernel approach has two main advantages: first, it is fast, as it is implemented using a novel two-layer random feature representation of the input message distributions; second, it has principled uncertainty estimates, and can be cheaply updated online, meaning it can request and incorporate new training data when it encounters inputs on which it is uncertain. In experiments, our approach is able to solve learning problems where a single message operator is required for multiple, substantially different data sets (logistic regression for a variety of classification problems), where it is essential to accurately assess uncertainty and to efficiently and robustly update the message operator. date: 2015-07-12 publisher: AUAI Press official_url: http://auai.org/uai2015/proceedings/papers/235.pdf vfaculties: VFLS oa_status: green full_text_type: pub language: eng primo: open primo_central: open_green verified: verified_manual elements_id: 1033814 lyricists_name: Gretton, Arthur lyricists_name: Jitkrittum, Wittawat lyricists_name: Lakshminarayanan, Balaji lyricists_name: Szabo, Zoltan lyricists_id: AGRET87 lyricists_id: JWITT76 lyricists_id: BLAKS90 lyricists_id: ZSZAB96 actors_name: Szabo, Zoltan actors_id: ZSZAB96 actors_role: owner full_text_status: public series: Conference on Uncertainty in Artificial Intelligence (UAI'15 ) publication: Uncertainty in Artificial Intelligence - Proceedings of the 31st Conference, UAI 2015 volume: 31 place_of_pub: Virginia, USA pagerange: 405-414 event_title: Thirty-First Conference on Uncertainty in Artificial Intelligence (UAI'15 ) institution: Conference on Uncertainty in Artificial Intelligence (UAI) book_title: Proceedings of the Thirty-First Conference on Uncertainty in Artificial Intelligence (UAI'15 ) editors_name: Meila, Marina editors_name: Heskes, Tom citation: Jitkrittum, W; Gretton, A; Heess, N; Eslami, SMA; Lakshminarayanan, B; Sejdinovic, D; Szabó, Z; (2015) Kernel-based just-in-time learning for passing expectation propagation messages. In: Meila, Marina and Heskes, Tom, (eds.) Proceedings of the Thirty-First Conference on Uncertainty in Artificial Intelligence (UAI'15 ). (pp. pp. 405-414). AUAI Press: Virginia, USA. Green open access document_url: https://discovery-pp.ucl.ac.uk/id/eprint/1470010/1/jitkrittum15kernel.pdf document_url: https://discovery-pp.ucl.ac.uk/id/eprint/1470010/6/jitkrittum15kernel_spotlight.pdf document_url: https://discovery-pp.ucl.ac.uk/id/eprint/1470010/11/jitkrittum15kernel_poster.pdf