4th Master Course on
Deep Learning for Artificial Intelligence
Universitat Politecnica de Catalunya
ETSETB TelecomBCN (Autumn 2020)
Previous editions: [2017] [2018] [2019] [All DL courses]
Deep learning technologies are at the core of the current revolution in artificial intelligence for multimedia data analysis. The convergence of large-scale annotated datasets and affordable GPU hardware has allowed the training of neural networks for data analysis tasks which were previously addressed with hand-crafted features. Architectures such as convolutional neural networks, recurrent neural networks or Q-nets for reinforcement learning have shaped a brand new scenario in signal processing. This course will cover the basic principles of deep learning from both an algorithmic and computational perspectives.
Instructors
Lecturers
Lab Instructors
Invited Guest
Lectures
Lectures - Mondays 2-4pm
Course will be divided in modules of 50 minutes (aproximately) covering the following topics. Slides & videos from previous edition are available here.
Neural Architectures
- (XG) The Perceptron [Slides 2019] [Video]
- (XG) Softmax Layer [Slides 2019] [Video]
- (XG) Multi-Layer Perceptron [Slides 2019] [Video]
- (VV) Convolutional Neural Networks (CNN) [Slides from AA2 2019]
- (VV) Neural Architectures from ImageNet [Slides 2019]
- (XG) CNN Visualization [Slides 2019] [Video 2018]
- (XG) Recurrent Neural Networks [Slides MCV 2020] [Video]
- (XG) Attention Mechanisms [Slides] [Video 2018]
- (XG) Transfomers
- (JR) Graph Neural Networks (GNN)
- (VV) Neural Architecture Search [Slides 2019]
- (RM) Compression [Slides 2018]
Training Deep Neural Networks
- (XG) Backpropagation [Slides 2019] [GIF] [Video]
- (JR) Loss functions [Slides 2019]
- (VV) Optimizers for Deep Learning [Slides 2018]
- (VV) Parameter initialization
- (VV) Normalization strategies
- (JR) Methodology [Slides 2019]
![[GIF]](https://github.com/telecombcn-dl/dlai-2019/raw/master/slides/dlai_2019_d02l2_backprop.gif){kind=link}
Learning Paradigms
- (XG) Machine Learning Basics: Supervised Learning [GSlides@upc.edu] [Video]
- (RM) Transfer learning & Domain Adaptation [Slides] [Video 2018]
- (RM) Incremental learning [Slides] [[Video 2018]][d06l1-video]
- (XG) Self-supervised Learning [Slides X-Europe 2020] [Video X-Europe 2020]
- (YK) Learning efficient representation for image and video understanding
- (XG) Self-supervised Audiovisual Learning [Slides] [Video]
- (XG) Generative Adversarial Networks (GANs) [Slides] [Video]
Labs
Labs - Thursdays 12-2pm
The course will contain guided hands on lab that will lead the students in their first steps in deep learning frameworks. Students must bring their own laptops to follow these labs.
- Lab01 A World of Tensors and Differentiable Computing Notebook Video
- Lab02 Automatic Differentiation with PyTorch Notebook Video
- Lab03 Linear regression Notebook
- Lab04 Multi-Layer Perceptrons Notebook
- Lab05 Convolutional Neural Networks Notebook
- Lab06 Fighting Overfitting Notebook
- Lab07 Transfer Learning Notebook
- Lab08 CNN Interpretability Notebook
- Lab09 Generative Adversarial Networks Notebook
- Lab10 Graph Neural Networks Notebook
- Lab11 Recurrent Neural Networks Notebook
Practical
Practical details
- Course on UPC Atenea
- Study Programs: Master MET & Master MATT at ETSETB TelecomBCN from the Universitat Politecnica de Catalunya.
- Course code and official guide: 230706 - DLAI
- ECTS credits: 5 ECTS
- Teaching language: English
- Capacity: 40 MSc students
Registration
Registration
Registration procedure depends on the student profile:
-
Master students at ETSETB: Follow the regular schedule from your academic office.
-
Bachelor (grau) students at ETSETB and CFIS students: Do not register to this course but to “Introduction to Deep Learning”, a lighter version of this course offered as a intensive Winter School between 22-28 January 2020. Check with your academic office.
-
MSc Students at UPC but not in ETSETB: Contact the your academic office and request being allowed to take this course. If accepted, contact ETSETB academic office and request more details.
-
Other students currently enroled in another academic program outside UPC: You can request attending to the seminar by sending your CV and motivation letter to the ETSETB academic office, clearly showing your previous knowledge in deep learning and their software frameworks. If accepted, you will need to cover the full cost of the course (around 720 euros, which correspond to 5 ECTS). You must already be legally eligible to attend to this course, we will not release any acceptance later for visas.
-
Other profiles (eg. industry members): This course is not addressed to this audience. If you are interested in this topic, we suggest you consider the postgraduate program in Artificial Intelligence with Deep Learning from UPC School. This postgraduate extends the dep learning contents offered at UPC TelecomBCN. Next edition starts in November 2019.
More courses
Previous & Parallel Courses
- Postgraduate Program in Deep Learning for Artificial Intelligence. UPC School. [online] or [onsite]
- Introduction to Deep Learning. UPC TelecomBCN. [2018] [2019]
- Deep Learning for Artificial Intelligence. UPC TelecomBCN. [2017] [2018] [2019]
- Deep and Reinforcement Learning. UPC TelecomBCN. [2020]
- Deep Learning for Computer Vision UPC TelecomBCN. [2016] [2017] [2018] [2019]
- Deep Learning for Speech and Language UPC TelecomBCN. [2017] [2018]
- Deep Learning for Video. Master in Computer Vision Barcelona. [2018] [2019] [2020]
- Multimodal Deep Learning. MMM 2019. [2019]
- Deep Learning for Multimedia. Insight Dublin City University 2017. [2017] [2018]
- Amaia Salvador and Santiago Pascual. “Hands on Keras and TensorFlow”. Persontyle 2017.
- Santiago Pascual. “RNN & GANs in PyTorch”. UPC TelecomBCN 2017.
Organizers
ETSETB Telecom BCN
Universitat Politècnica de Catalunya
Sponsors
[Online edition](https://www.talent.upc.edu/ing/estudis/formacio/curs/310401/postgraduate-course-artificial-intelligence-deep-learning/)
[Onsite edition](https://www.talent.upc.edu/ing/estudis/formacio/curs/310400/postgraduate-course-artificial-intelligence-deep-learning/)