The ancient history of deep learning goes back to the 50’s with Rosenblatt’s perceptron, but the real revolution started around 15 years ago (Hinton, Lecun). Deep learning methods now dominate most of the machine learning and AI solutions and outperform classical techniques and even humans on some tasks. However, for those who believe classical methods lost the battle and are no longer relevant, Dorit will prove you wrong!
Classical machine learning has many advantages over deep learning methods in terms of the volumes of required data, interpretability, computational resources and more. In this talk, Dorit will concentrate on the advantages of classical classification methods compared to deep learning classifiers. She will provide concrete examples of problems that can be tackled successfully by classical machine learning. Dorit will also provide guidelines based on her experience of over 15 years in the field on how to choose the best solution whether it is classical, deep or hybrid.