| Abstract： |
We demonstrate why a new computing paradigm, called Data Mechanics, and an integrative inferential concept, called pattern inference, are needed to constitute a foundation for learning-from-data in complex systems. We exclusively illustrate our computational developments in a winemaking-to-winetasting system. Such an example complex system begins with a water stress experiment on a Cabernet Sauvignon vineyard located at Esparto, CA, then goes through three keystone phases of winemaking performed in UC Davis Pilot winery, and ends with a sensory study of winetasting conducted in Robert Mondavi Institute for Wine and Food Sciences. Data derived from the entire system is represented by longitudinal bipartite networks. The biological questions are: 1) Can effects of water stress regimes affect characters of bottled wines at the final phase of winemaking? 2) What are computable systemic understanding of winemaking? 3) How to quantitatively evaluate winetasting? We give an overview on how to computationally resolve all these holistic questions. Our demonstrations on why statistics as a discipline is not ready to address these questions are focused on the lack of systemic robustness/sensitivity considerations in dimensional reduction and model selection methodologies. |
