The School of Mathematical and Computational Sciences welcomes the campus community to public research and teaching presentations by Dr. Arne Dankers, candidate for the tenure track position in Data Science. All are welcome to attend. 

Research Talk Title: "Data Driven Modelling of Dynamic Networks". 10am-11am

Abstract: Data is not information, but there is a lot of information encoded into data! The trick of course is to design methods that are able to extract the information from the data. Imagine being able to discover the inner workings of cells (such as how they process sugar), simply by measuring the concentration of various chemicals over a time period after feeding the cell some sugar. Imagine monitoring and diagnosing the inner components of a machine based on sensor data. Imagine discovering the causal structure of an economic system by processing financial data. Using data we can peer into the structure and inner workings of complicated systems leading to scientific discoveries, effective monitoring of engineering systems, and a deeper understanding of economic, social, ecological systems among many others! In this talk I present a framework that enables the user to estimate differential equations that connect the measured variables in a data set. I present some interesting problems that must be addressed in order to improve the framework. In addition I present two areas where I have applied this framework: geology and pipeline monitoring. 

Teaching Talk Title: "Model Selection" 12:30pm-1:30pm

Abstract: The model selection workflow is one of the main occupations of a data scientist. It is the procedure of collecting data, choosing a model set and then picking the "best" model. It is quite likely that the first model will not pass the validation test, and so the procedure must be repeated (many times). In this lecture I focus on the validation criterion portion of the model selection workflow. I will present k-fold cross validation as a tool to select a model structure. I will point to strategies that help to focus the modelling effort to the most beneficial action (collect more data? change the model set? choose a different selection criterion?). Code is implemented in python. 

Date and Time: April 27th, Research Talk 10am-11am/Teaching Talk 12:30pm-1:30pm

Via Zoom

https://zoom.us/j/93748112808?pwd=NE5WTzMrN1pQREFKcVhvcU4wL3loUT09

Meeting ID: 937 4811 2808
Passcode: 936409