Cognitive: Foundation and mathematical formalization of the learning and generalization concept and its importance in the computational intelligence field. Comparative conception of the basic neural network and deep learning architectures. Training principles and learning abilities of the learning machines. Exposition to the diverse applications of neural networks and deep learning in the fields of pattern recognition, function approximation, data mining and information retrieval.
Skills: Promoting analytic and implementation skills in pattern recognition and function approximation using neural networks deep learning and implementation of the corresponding deep neural network training algorithms. Use of deep learning libraries for real problems. Promoting programming skills. Ability to complete a learning problem by implementing a neural network algorithm, running experiments on real data, writing a short paper and presenting the results.
Course Content (Syllabus)
Neural Nwtworks and Deep Learning introduction. Back-propagation. Deep Autoencoders and representation learning. Deep Boltzmann Machines. Convolutional Neural Networks. Reccurent Neural Networks. Generative adversarial networks. Deep Reinforcement Learning. Knowledge transfer. Optimization, regularization, over-training and generalization in deep learning. DL model architecture design and training. DL libraries and frameworks. Computational complexity of DL models and use of parallel programming in GPUs. DL for embedded systems. Applications of DL for data analysis, classification, clustering and retreival. Applications on text, audio, image and video. Applications on autonomous systems and robotics.
Course Bibliography (Eudoxus)
1. ΤΕΧΝΗΤΑ ΝΕΥΡΩΝΙΚΑ ΔΙΚΤΥΑ, Κ. ΔΙΑΜΑΝΤΑΡΑΣ, ΚΛΕΙΔΑΡΙΘΜΟΣ, 2007, ΑΘΗΝΑ
2. ΝΕΥΡΩΝΙΚΑ ΔΙΚΤΥΑ ΚΑΙ ΜΗΧΑΝΙΚΗ ΜΑΘΗΣΗ, 3η ΕΚΔΟΣΗ, S. Haykin, Παπασωτηρίου, 2010, Αθήνα.
3. Διαφάνειες και υλικό σε ηλεκτρονική μορφή
4. Free E-books στα Αγγλικά