The lecture will introduce the topic of Artificial Neural Networks. We will
start by using single and multilayer perceptron networks to solve
classification and regression problems. Next it will be shown how networks can
help to reduce the dimensionality of the data and detect relations present in
the data through the introduction of Self Organizing Maps. Final lectures will
concentrate on the Deep Learning approach to build multi-layered networks one
layer at a time. Topics related to Neural Networks, such as Support Vector
Machines and matrix factorizations, such as PCA and ICA will be introduced to
provide another solution to the problems discussed. The lecture will be
accompanied by computer exercises that will teach the practical aspects of
Neural Networks implementation and usage. **Course prerequisites:** it will be
beneficial to know the basics of statistics, numerical analysis, and have a
good grasp on programming in Matlab or Python+numpy. **Tentative curriculum:**
1. Introduction to learning from data, learning machines, and artificial neural networks.
2. Single and multi-layer perceptron networks.
3. Artificial Neural Network training methods, both online and batch oriented.
4. Support Vector Machines.
5. Introduction to unsupervised learning.
6. Unsupervised methods: Matrix factorizations (PCA, ICA, NMF), Self-organizing maps.
7. Radial Basis Function networks: concepts, applications and design methods.
8. Current trends in Artificial Neural Networks: Deep Learning.
Fri 12:15-2pm, you must book a slot at https://calendar.google.com/calendar/selfsched?sstoken=UUtYVm5mOVkxUVRNfGRlZmF1bHR8ZmJjZmRiMzU4ODA0NDAyYTYzOTZiZjNjY2Y2ZDQwODk