PhD Sebastian Mair har sin docentföreläsning måndag 9 februari. Forskarutbildningsämnet är datalogi och föreläsningens titel är:
Data-Efficient Machine Learning
Abstract:
Recent progress in machine learning has been largely driven by scale. Increasing data sizes and model capacity have enabled major performance gains, but at the cost of growing storage requirements, computational demand, and energy consumption. This lecture introduces the principle of data frugality: achieving strong predictive performance while using only as much data as necessary. Drawing on our research on data subset selection across multiple learning settings, we show how carefully chosen training data can reduce storage requirements, training time, and the carbon footprint, with negligible performance drops. The lecture concludes with an outlook on emerging research directions and open challenges in data-efficient machine learning.
Föreläsningen hålls på engelska.
Alla är välkomna!
