Invited Speakers

Professor Paul Cohen

Paul Cohen is a professor of Information Science at the University of Arizona and the manager of three programs at the Defense Advanced Research Projects Agency (DARPA), part of the US Department of Defense. At Arizona, Cohen served as head of Computer Science and founding director of the university's School of Information: Science, Technology and Arts. He holds a PhD in Computer Science and Psychology from Stanford University, a Master of Science degree in Psychology from the University of Californai, Los Angeles, and a Bachelor of Science degree in Psychology from the University of California, San Diego.

Cohen's research interests span artificial intelligence and include machine learning, language, vision, semantic technology, data analysis, information theory and education informatics. His programs at DARPA are concerned with technologies to help humans model and manage very complicated systems, such as cancer molecular signalig pathways and the food-water-energy nexus.

Mining for Causal Results

Science proceeds from description and classification to prediction, then strives for explanation. Classification and prediction do not require causal knowledge, explanation does. Thus, the most valuable scientific results establish cause-effect relationships. While these relationships may be implicit in data, they are explicitly stated in scientific publications. Extracting causal relationships from data is difficult, whereas extracting explicit causal assertions from scientific papers is relatively easy; which makes today's preoccupation with Big Data at least puzzling and perhaps wrong-headed. Hints from DARPA's Big Mechanism program suggest that domain-specific causal knowledge in scientific papers can be mined by machines using domain-independent methods, which would make these the methods of choice for scientists who prefer causal assertions to vast amounts of data that might or might not harbor causal assertions. A "third way" envisions data as a source of causal assertions that scientists might otherwise miss, and as evidence pertaining causal assertions mined from scientific literature. Of course, shifting the unit of analysis from data to causal assertions raises many research issues, including the problem of integrating published assertions, which tend to be fragmentary, into larger models or theories; and the need to maintain the experimental contexts of these assertions; and the likelihood that at least some assertions are wrong. The talk will illustrate these points with concrete examples of mining causal assertions about cancer biology.

Dr A. Aldo Faisal

Dr Aldo Faisal is Associate Professor for Neurotechnology at Imperial College London (Dept. of Bioengineering & Dept. of Computing) and Director of the Behaviour Analytics Lab at the Data Science Institute (London). He is Associate Group Head at the MRC London Institute for Medical Sciences and affiliated faculty at the Gatsby Computational Neuroscience Unit (University College London). In 2009 he was appointed to Imperial College London to develop a human-based Neurotechnology research area, where he established his lab. Dr Faisal's combines cross-disciplinary computational and experimental approaches to investigate how the brain and its neural circuits are designed to learn and control goal-directed movements. The neuroscientific findings enable the targeted development of novel technology for clinical and research applications (Neurotechnology) for a variety of neurological/motor disorders and amputees. Aldo serves on the editorial board of Nature Scientific Data and PLoS Computational Biology, and at international conferences (general/program/area chair) such as Neurotechnix, various IEEE conferences and NIPS. He was recently elected to the World Economic Forum's Council on the future of Neurotechnologies and Brain Science.

Reverse engineering human decision making from high resolution analysis of behaviour

Our research focuses the computational analysis of human behaviour using data-driven analysis and prediction, including the development of methods & algorithms to move from Data to Knowledge. Our key Goals are to predict human behaviour from ubiquitous sensors & digital data and so to predict and evaluate human performance, as well as inferring internal or cognitive state of individuals from behavioural dynamics. We will illustrate these goals with current examples in our Human Ethome Project as well as supporting medical decision making in the clinic with pervasive sensing.

Dr Niels Peek

Niels Peek (MSc, PhD) is Associate Professor in Health Informatics at the University of Manchester. With a background in Computer Science and Artificial Intelligence, his research focuses on data-driven informatics methods for healthcare quality improvement, data mining for healthcare, predictive models, and clinical computerised decision support. He is director of the Greater Manchester Connected Health City, which is part of the B#20M "Health North" investment to establish a learning health system in the North of England. Dr. Peek has co-authored more than 150 peer-reviewed, scientific publications. Previously based at the University of Amsterdam, the Netherlands, he led the "CARDSS" initiative, a collaboration between academic partners, professional and patient organisations in cardiac rehabilitation, and industry partners which led to the introduction of computerised decision support in 40 Dutch hospitals and national quality standards for cardiac rehabilitation. He is the President of the European Society of Artificial Intelligence in Medicine, editorial board member of the Journal of the American Medical Informatics Association and Artificial Intelligence in Medicine, and co-chair of the Scientific Programme Committee of MEDINFO-2017, the 16th World Congress on Health and Biomedical Informatics.

Opportunities and Challenges of Learning Health Systems

Health systems worldwide are under pressure to deliver better care for more people from fewer resources. The global economic crisis has shrunk the resources available for healthcare but the growth in demand for care services continues unabated. "Learning Health Systems" is a novel health informatics paradigm that blends quality improvement methods with data science. The goal is to create integrated healthcare systems which harness the power of data and analytics to learn from every patient, and feed the knowledge of "what works best" back to clinicians, patients, and other stakeholders of the healthcare system, to create cycles of rapid improvement. In this talk we dissect the new paradigm and explore its opportunities and challenges for data scientists.