Re-representing behavior models with machine learning methods: opportunities for logic programming

On 28 January at 1pm, Professor Francisco Camara Pereira from DTU Management will give a talk at DTU Compute, in room 240, building 324 (the lunch room). There are clear connections between Francisco’s work in AI and what we do at DTU Compute, so one of the purposes of his talk is to get inspiration for possible future collaboration. Everybody is very welcome, and I hope many of you will decide to come. Details below.

Modeling and predicting human behavior has been a lively research topic for decades. There is an entire established area, generally called behavior econometrics, that dedicates particularly to how people make choices, and how to use such knowledge to guide and inform interventional changes (e.g. policy making, infrastructure investments, marketing). Both from a research and practice stand-point, knowledge and data representation have followed the same principles for decades, essentially based on statistics and econometrics. However, such representations are arguably not ideal in many current challenges (e.g. social influence in behavior, hierarchical decision making, spatial and temporal causality).

In this presentation, I will show some recent work that tries to address some of these newly recognized challenges, using Machine Learning research, such as generative modeling, neural network embeddings, or context free grammars. My main goal is, however, to present and discuss the potential of a logic programming based direction, which I believe is very exciting.

Tidspunkt

tir 28 jan 20
13:00 - 14:30

Hvor

324/240 at DTU Compute
(the lunch room)