Activity: Talk or presentation types › Contributed talk › Scientific
Description
Deep Learning emerged nowadays as a successful problem solver from very large scale industrial applications used by important players on the market (e.g. Google, Microsoft) to advances in state-of-the art theoretical artificial intelligence (e.g. Deep Q-Network from Google Deep Mind). These are just the tip of the iceberg, but there is a much larger number of problems which were solved using deep learning such as: multivariate time series classification and regression, computer vision, transfer learning, quality of experience, and so on. Still, deep learning is far away from being perfect and there are a number of limitations which constrain its strength and applicability such as suboptimal learning algorithms, high computational costs, and lack of mathematical proves. Thus, this talk will present arguments in the favour but also against deep learning, and it will end up with a series of open questions in deep learning and in its possible application domains.