Friday, September 27th - h 2:00 p.m.
Seminars Room, NICO
Neuroscience Institute Cavalieri Ottolenghi
Regione Gonzole 10, Orbassano (TO)
Solutions and a problem
The data-driven learning paradigm has replaced rule-based reasoning as the de-facto modern approach to artificial intelligence, thanks to new algorithms and the unprecedented availability of data and computing power.
Amongst the variety of real-life applications, computer vision is arguably the domain that was impacted the most. Deep convolutional neural networks nowadays allow to automate a number of vision based tasks in a way that was unthinkable a few years ago.
However, there is a trade-off between model complexity and its explainability, and deep learning falls on the former end of the spectrum. We are now able to train accurate models that are intrinsically unable to provide a human-understandable explanation of their internal decision processes. The field of explainable Artificial Intelligence tries to peek into the black-box of modern artificial intelligence. Machine learning projects on urban, healthcare, energy, finance, and satellite data.
Host: Annalisa Buffo