Dr. Igor Mapelli will give a seminar on “Archaeological Data Analysis in the Era of Big Data” on October 30 at 4 pm. The abstract of his talk and his short bio is shared below.

Bio:
Igor Mapelli is a postdoctoral researcher at the Biology Department of the Middle East Technical University (METU). Igor completed his Ph.D. studies in medical informatics at METU in early 2019 during the course of which he worked on the electrophysiology of the human brain specializing in visual working memory. Before moving to Turkey, he earned his MSc in computer science from the University of Turin in Italy where he had concentrated majorly on data mining and artificial intelligence. His research interests lie in the areas of neural oscillations and machine learning, ranging from more theoretical aspects of human electrophysiology to practical applications such as brain-computer interfaces. He has collaborated actively on projects in few other disciplines which contributed to consolidating his data analysis expertise. Currently, in his postdoctoral position, he is applying machine learning and network analysis techniques on archaeological and biological data to help studying the evolution of mobile hunter-gatherer societies into sedentary village farmers.

Abstract:
Archaeological Data Analysis in the Era of Big Data The rapid growth of available data across the whole scientific disciplines opened novel opportunities with regard to both research questions and analytical methodologies that rely on machine learning for data-driven discovery. In this respect, the field of archaeology is no exception and in addition to the rise in the quantity of material culture and historical data, the heterogeneity of data is also increasing, thanks to the contribution from fields such as climate science, genetics, and medicine. Within this context, we will examine how statistical coefficients of relatedness (i.e., kinship coefficient θ and Cotterman coefficients k0, k1, k2) aiming at the characterization of genetic kinship levels among ancient individuals cannot reliably evaluate all cases. Thus, alternative approaches, from classical machine learning techniques to more recent deep learning approaches, are needed. Furthermore, we will present how network analyses – based on Kulczynski-2 similarity measures – may be applied over material culture data to study properties and relationships between geographical sites and across different time periods.