Estimation and learning in a Big Data context
Ciclo de Conferências 2019/2020 - Tecnologia, Empresa e Sociedade
Abstract
When learning from real-world data we face large, space-time dependent datasets with missing entries and with different types: numerical, categorical, ordinal. Our digital life increasingly requires learning and estimation algorithms for streaming, heterogeneous, and sometimes private data, in useful time, i.e., running with complexity that scales linearly with the size of data. In this talk I will cover some of my work in estimation and learning methods for Big Data, based on ongoing projects and partnerships, with an eye on the greater good
Speaker
Cláudia Soares (IST/UL; DEISI/ULHT)
Claudia Soares works in Big Data and distributed processing at ISR Lisboa, IST, where she is an Inv. Ass. Professor. She has also an appointment with U. Lusófona de Humanidades e Tecnologias. She strongly believes in developing research with academia and industry, exchanging ideas and methods for effective, simple, and scalable algorithms — and for solving real problems. Her research delves on developing a generic framework for learning with large-scale, heterogeneous, and space-time dependent data. For this work, Dr. Soares developed both parallel and stochastic algorithms, where nodes compute asynchronously, mostly with outdated data.
Approaching real-world problems resulted in successful proposals and publications in top venues. She has been collaborating with companies to address problems that are both real and aligned with her scientific interests. In this context, she worked with national and international industries like NOS, Nomad-Tech (CP), TAP, Thales, 3Lateral (Epic games), uRoboptics, and developed academic collaborations with International partners like the TU/e, U. Milano, and U. of Novi Sad. Dr. Soares has a key role in funded projects of the national agency (FCT) in Data Science and AI for predicting ECU admissions and EU funded projects in AI, namely AI4EU -- European AI On Demand Platform and Ecosystem, and a European Doctoral program, BIGMATH, for BIG data challenges for MATHematics.