Publications




Journals

Beyond Fano’s Inequality: Bounds on the Optimal F-Score, BER, and Cost-Sensitive Risk and Their Implications [PDF]
Ming-Jie Zhao, Narayanan Edakunni, Adam Pocock, and Gavin Brown
Journal of Machine Learning Research (JMLR). Volume 14, Pages 1033-1090, 2013.

Conditional Likelihood Maximisation: A Unifying Framework for Information Theoretic Feature Selection [PDF][Poster]
Gavin Brown, Adam Pocock, Ming-Jie Zhao, and Mikel Luján
Journal of Machine Learning Research (JMLR). Volume 13, Pages 27-66, 2012.
The paper is available here and the code for the various feature selection algorithms is on mloss or locally hosted here.

Conferences

Minimally-Constrained Multilingual Word Embeddings via Artificial-Code Switching   [PDF]
Michael Wick, Pallika Kanani, Adam Pocock
Thirtieth AAAI conference on Artificial Intelligence (AAAI). Phoenix, February 2016.

A Scalable Implementation of Information Theoretic Feature Selection for High Dimensional Data   [PDF]
Anthony Kleerekoper, Michael Pappas, Adam Pocock, Gavin Brown, and Mikel Luján
IEEE Intl Conference on Big Data. Santa Clara, October 2015.

Augur: Data-Parallel Probabilistic Modeling   [PDF] [Paper page]
Jean-Baptiste Tristan, Daniel Huang, Joseph Tassarotti, Adam Pocock, Stephen Green, Guy Steele
Neural Information Processing Systems (NIPS). Montreal, December 2014.
Published in Advances in Neural Information Processing Systems, Volume 27, Pages 2600-2608.

Informative Priors for Markov Blanket Discovery   [PDF] [Supplementary Material] [Poster]
Adam Pocock, Mikel Luján, and Gavin Brown
15th Intl Conference on Artificial Intelligence and Statistics (AISTATS). La Palma, April 2012.
Published in JMLR W&CP, Volume 22, Pages 905-913.

Online Nonstationary Boosting  [PDF]
Adam Pocock, Paraskevas Yiapanis, Jeremy Singer, Mikel Luján, and Gavin Brown
9th Intl Workshop on Multiple Classifier Systems (MCS). Cairo, April 2010.
doi: 10.1007/978-3-642-12127-2_21
A Java implementation of Online Boosting and ONSBoost is found here.
This paper was updated on 05-02-2017 to correct an issue with Algorithm 1. The original version set h prime to arg max instead of arg min.

Workshops

Minimally-Constrained Multilingual Word Embeddings via Artificial-Code Switching   [PDF]
Michael Wick, Pallika Kanani, Adam Pocock
Multi-task and Transfer Learning workshop, NIPS. Montreal, December 2015.

Attribute Extraction from Noisy Text Using Character-based Sequence Tagging Models   [PDF]
Pallika Kanani, Michael Wick, Adam Pocock
Machine Learning for eCommerce workshop, NIPS. Montreal, December 2015.

PhD Thesis

Feature Selection via Joint Likelihood  [PDF]
Adam Pocock
Supervisors: Gavin Brown, Mikel Luján
University of Manchester. August 2012.
Awarded the 2013 BCS Distinguished Dissertation prize.

MSc Thesis

Feature Selection using Information Theoretic Techniques  [PDF]
Adam Pocock
Supervisor: Gavin Brown
University of Manchester. September 2008.