ABSTRACT / One of the most important tasks in data science is the quantification of dissimilarity within a (possibly dynamically changing) system. We use dissimilarity measures when recognizing temporal or spatial patterns like trends or cliques. Important findings greatly depend on the definition of dissimilarity (or distance) measure. We frequently face the dilemma what metrics to use when exploring the dynamics of the system, while admitting that not all commonly used such metrics satisfy the triangle-inequality.
We show two dynamic systems where the dissimilarity concept of Bray-Curtis, commonly used in numerical ecology, is useful to recognize temporal and spatial patterns. The data are from reference databases (US Patent Office and Microsoft Academic Graph). The patents, as well as the publications, are assigned to technological categories. The interactions between these categories form a network, where the edges link the categories by achievements belonging to one category and referring to other achievements belonging to other categories. This mechanism generates a weighted, directed, dynamically changing network. While its structure is stable, its internal dynamics, at least within the Bray-Curtis space, is lively and shows a consistent pattern. This raises the question whether this is primarily due to the chosen metrics, or due to some intrinsic features of how we refer to each other’s achievements in our Intellectual Ecological Space.
Our aim is not to show solutions to these questions, rather to discuss them and try to convince the audience that this is an exciting research topic potentially leading to new insights in the dynamics how our knowledge is being accumulated.
BIO / József Baranyi is a Hungarian-British mathematician, who worked for the Institute of Food Research in the United Kingdom for 26 years, leading the Computational Microbiology Research Group there. He was also a Visiting Professor at the Physics Department of Imperial College in London through 2012-2019. After retirement in the UK, he became a Scientific Advisor at the University of Debrecen, now leading the Predictive Nutrition Research Group there.
He has coordinated or participated in several international projects; authored or co-authored ca 100 research papers, book chapters and other scientific communications, with a total citation of more than 6000 (Scopus, 2021). The Baranyi-model on bacterial growth is one of the most frequently quoted models in predictive microbiology. He was the Statistical Advisor of the Journal of Applied Microbiology for 14 years and Editorial Board member at the Applied and Environmental Microbiology for 15 years, for which he received the “Distinguished Service Award” of the American Society for Microbiology. He is also an elected member of the International Academy of Food Science and Technology, the prime advisory body of the International Union of Food Science and Technology.
In 2018, he became a Doctor Honoris Causa and Private Professor of the Szent-István University (today Hungarian University of Agriculture and Life Sciences) in Budapest.
Ákos Jóźwiak is the research director of the Digital Food Institute at the University of Veterinary Medicine, Budapest. Before, he worked for the National Food Chain Safety Office (NÉBIH) and its predecessors for 15 years in various positions.
He is a member of the European Food Safety Authority (EFSA) Advisory Forum and the EFSA Emerging Risk Exchange Network. He also chairs the EFSA Advisory Group on Data.
In his research activities, he focuses on developing and applying new methods for improving the effectiveness of the controls of the food system. Within this domain, his main research areas are (1) applying computational science methods for identifying food system emerging risks; (2) applying data science for optimizing food production; (3) determining the economic burden of foodborne diseases and applying health technology assessment methods for food chain safety decision making.