This meeting marks the launch of Durham's new Institute for Data Science. It aims to give researchers at the University the opportunity to present their interests in Data Science, to network, and to get to know each other.
Over the centuries, scientists have sought to describe natural phenomena with mechanistic models. In the last century, we have also made tremendous progress in building powerful non-mechanistic models, from linear regression to random forests and neural networks. However, although data are nowadays abundant in many areas of science and engineering, models often remain elusive. I will introduce the toolbox of sparse regression and illustrate how it can extract parsimonious models from data. I will give the examples of charging electric vehicles and automating the discovery of governing equations from data.
Working with image data provided by the Durham University Oriental Museum, we sought to test how well a data set created from a museum's digital archives could be used for broad artefact type classification using a CNN implemented in PyTorch, as well as considering potential ways in which computer vision and object recognition could be used to support efforts in heritage protection.
This talk is based on research conducted in collaboration with the United Nations and demonstrates the ability with which AI can be used to generate text easily, and at a low cost. The implications of this to societal peace and stability will be explored, and recommendations for potential steps towards the mitigation of negative consequences given.