22nd January - 09:30 (Link: https://videoconf-colibri.zoom.us/j/87292833142) 

In the era of Big Data, IOT, precision medicine, automatic classification plays a fundamental role. We need algorithms that are able to classify images, identify malfunctions, recognize the pattern of a disease ... the examples are innumerable. So we will talk about automatic classification and show how, behind the algorithms, we have interesting optimization problems to solve.

Machine Learning pipelines that are in production are prepared to automatically apply inference on a specific data distribution. However, many are the use cases where this distribution may change over time, thus producing a negative impact on model performance. How can we detect this shift? In this presentation we explore a potential solution to this problem in the context of detecting transaction fraud.

__________________________________________________

Paula Amaral is an auxiliary professor in the Department of Mathematics at FCT NOVA and a member of the Center for Mathematics and Applications (CMA) at Universidade Nova de Lisboa. She has a degree and Master in Statistics and Operational Research from the University of Lisbon and a PhD in Mathematics from the New University of Lisbon. Her research area is in Optimization, Modeling and Optimization for Machine Learning having published several papers. She has supervised several internships in industrial and consulting companies such as Telepac, Águas de Portugal, CTT, GALP and supervised master’s students from national (FCT UNL) and international schools (Sapienza, Rome). She is a member of the Board of the Portuguese Mathematical Network for Industry and Innovation (PT-MATHS-IN), a member of the Executive Committee of the CMA and a member of the Board of Europt (European Group for Continuous Optimization).

Simão Gonçalves is a Master's student finalist at NOVA School of Science and Technology (FCT NOVA). He first completed the initial 3 years of the Electrical & Computer Engineering Integrated Masters in 2018. Then, later that year he began a Master's course in Analysis and Engineering of Big Data and his thesis is focused on modelling patient trajectory for primary care data.