Michigan Institute for Data Science (MIDAS) Seminar Series: “The Enigmatic Kime: Time Complexity in Data Science”

Ivo Dinov, PhD, Michigan Alzheimer’s Disease Core Center Data Core Co-leader & Biostatistics and Data Management Core Co-Lead for the Udall Center for Parkinson’s Disease Research, presented at the Michigan Institute for Data Science (MIDAS) Seminar Series The Enigmatic Kime: Time Complexity in Data Science on Friday, September 21, 2018.



Abstract: We provided a constructive definition of “Big Biomedical/Health Data” and provided examples of the challenges, algorithms, processes, and tools necessary to manage, aggregate, harmonize, process, and interpret such data. In data science, time complexity frequently manifests as sampling incongruency, heterogeneous scales, and intricate dependencies. We presented the concept of 2D complex-time (kime) and illustrate how the kime-order (time) and kime-direction (phase) affect advanced predictive analytics and scientific inference based on Big Biomedical Data. Kime-representation solves the unidirectional arrows of time problems, e.g., psychological arrow of time reflects the irrevocable past to future flow and thermodynamic arrow of time reflecting the relentless growth of entropy. Albeit kime-phase angles may not always be directly observable, we illustrated how they can be estimated and used to improve the resulting space-kime modeling, trend forecasting, and predictive data analytics. Simulated data, clinical observations (e.g., neurodegenerative disorders), and multisource census-like datasets (e.g., UK Biobank) were used to demonstrate time-complexity and inferential-uncertainty.