Teaching Plan Information

5.1. Methodological overview

The methodology followed in this course is oriented towards achievement of the learning objectives. A wide range of teaching and learning tasks are implemented, such as

• Lectures. 
• Talks from experts. 
• Lab sessions. 
• Practice sessions.
• Tutorials. 
• Autonomous work and study.
• Assignment.
• Assessment. Oral presentation of the practical assignment.

5.2. Learning tasks

The course has 3 ECTS, that is equivalent to approximately 75 hours of student work (35 of lectures and practice sessions and 40 hours of autonomous work). 

5.3. Syllabus

The course will address the following topics:

1. Introduction.
2. Deep Neural Networks.
3. Training of Deep Neural Networks. Backpropagation.
4. Deep Convolutional Networks.
5. Practical aspects of deep neural networks (Numerical gradient, overfitting, regularization, activation functions, stochastic gradient descent).
6. Application cases: Natural Language Processing and Visual Classification.
7. Hardware and software tools for deep learning.
8. Sequential Learning.
9. Approximate Nearest Neighbour.

5.4. Course planning and calendar

The course takes place in the Rio Ebro Campus. Sessions are organized in

5.5. Bibliography and recommended resources

• Bengio, Yoshua. Deep Learning / Yoshua Bengio, Ian Goodfellow and Aaron Courville. . . MIT Press (In preparation) [Comentario del profesor: preliminar versions of the chapers here http://www.deeplearningbook.org/] [Obs. docente: preliminar versions of the chapers here http://www.deeplearningbook.org/]
• Watt, Jeremy. Machine Learning Refined / Jeremy Watt, Reza Borhani, Aggelos Katsaggelos Cambridge University Press,2016
• preliminar versions of the chapers here (Yoshua Bengio, Ian Goodfellow and Aaron Courville. Deep Learning. MIT Press. In preparation )
• Yann LeCun, Yoshua Bengio and Geoffrey Hinton. Deep Learning. Nature 521, nº 7553 (2015): 436:444 (Visible para usuarios UNIZAR)