O Departamento de Engenharia Informática (DEI) do Instituto Superior de Engenharia do Porto (ISEP), numa iniciativa conjunta com o Mestrado em Engenharia de Inteligência Artificial (MEIA), convida à participação em mais uma AI Master Class, desta vez sobre o tema "Towards more dynamic and efficient Machine Learning pipelines", que terá lugar no dia 13 de dezembro, pelas 19h00, no ISEP.
O evento será presencial, na sala B202.
At a time in which machine learning algorithms have never been so powerful, and when we have unprecedented access to computational resources, it feels rather contradictory that we still face some of the toughest technical and scientific challenges of the field to date. This can be explained by several factors. Data are now dynamic, streaming rather than batch, with their statistical properties changing over time. So, models must be updated regularly. But data are also growing, so training new models takes more time, often too much. And how to pick the best model/configuration, in a time in which the possibilities are virtually infinite, and in which environmental concerns and legislation push towards a more judicious use of computational resources? These challenges are the starting point for this Masterclass, in which different solutions, developed in the context of several funded research projects, will be presented. Specifically, we will discuss how approaches such as meta-learning or optimization can be used to improve current machine learning pipelines.