André M. Carrington

Ph.D M.Math P.Eng

acarringtontoh.ca

 

Measuring predictive accuracy (AUC, etc) in parts of an ROC curve or groups of predicted risk and an interpretation of AUC for individuals not just pairs.
Carrington AM, Manuel DG, Fieguth PW, Ramsay T, Osmani V, Wernly B, Bennett C, Hawken S, McInnes M, Magwood O, Sheikh Y, Holzinger A. Deep ROC Analysis and AUC as Balanced Average Accuracy for Improved Classifier Audit, Selection and Explanation. IEEE Transactions on Pattern Analysis and Machine Intelligence, Early Access, Jan 25, 2022. doi:10.1109/TPAMI.2022.3145392 code, presentation, website

Updated laws and governance can greatly accelerate and broaden research access to health data with consent for innovation.
Carrington AM, Manuel DG, Bennett C. FAIR Access to Personal Health Information in Private and Public COVID-19 Health Applications. Authorea. October 28, 2020. doi:10.22541/au.160391056.64115187/v1

Machine learning and statistics predict syncope in emergency rooms.
Grant L, Joo P, Carrington A, Nemnom M, Thiruganasambandamoorthy V. Risk-stratification of emergency department syncope by artificial intelligence using machine learning: human, statistics or machine in CAEP/ACMU 2020 Scientific Abstracts June 1st - 4th, 2020, Ottawa, Ontario. CJEM, 22(S1), S1-S1. doi:10.1017/cem.2020.44

The proper generalization of the AUC and C statistic to measure parts of an ROC curve or plot, since the partial AUC is misleading. We define the concordant partial AUC and partial C statistic.
Carrington AM, Fieguth PW, Qazi H, Holzinger A, Chen HH, Mayr F and Manuel DG. A new concordant partial AUC and partial c statistic for imbalanced data in the evaluation of machine learning algorithms, BMC Medical Informatics and Decision Making 20, 4 (2020) doi:10.1186/s12911-019-1014-6. errata, code, presentation 2021, presentation 2020

Measuring our understanding of the output, explanations and causality of an AI system for trust, knowledge integration and optimal effect.
Holzinger A, Carrington AM, Müller H. Measuring the Quality of Explanations: The System Causability Scale (SCS). Künstliche Intelligenz (German Journal of Artificial Intelligence), 34 (2), 2020. doi:10.1007/s13218-020-00636-z

Some AI methods, like support vector machines make decisions in an infinite dimensional space--which is an opaque black box to clinicians. For tabular data, we achieve better performance with a new transparent method and new measures of transparency. We explore a hybrid of data modeling & machine learning.
Carrington AM. Kernel methods and measures for classification with transparency, interpretability and accuracy in health care. (Ph.D dissertation). University of Waterloo. UWSpace. 2018.

Measuring the transparency and interpretability of AI models and kernels.
Carrington AM, Fieguth PW, and Chen HH. Measures of model interpretability for model selection. In International Cross-Domain Conference for Machine Learning and Knowledge Extraction, Springer, 2018. preprint (open access), presentation

A new transparent similarity function (kernel) classifies more accurately than standard kernels and is safer than its likeness among them: the sigmoid kernel.
Carrington AM, Fieguth P, Chen H. A New Mercer Sigmoid Kernel for Clinical Data Classification. In Engineering in Medicine and Biology Society (EMBC), 2014. 36th Annual International Conference of the IEEE (pp. 6397-6401). IEEE. preprint (open access), presentation


Previous Career

Analysis of blockchain, homomorphic encryption, differential privacy and other technologies in research.
Carrington AM, Zhu M. Privacy and security technology in secondary use of health information. Technical Report (section). University of Waterloo, 2019.

Bojicic S, Paré D, Caron R, Carrington AM, and MacKenzie L, et al. Canada Health Infoway: Transport Layer Interoperability Specification - Canadian Draft for Use, 2011. related presentation (TLI security)

Mussi J, Carrington AM, et al. Electronic medical record specification. Canada Health Infoway, 2010.

Petersen I, Carrington AM, Banghart J, et al. Center for Internet Security: Wireless Security benchmark v1.0, 2005. Linksys addendum


Google Scholar Profile

My Google scholar citations