What are you looking for?
Theory and labs in English.
Documentation in English
B5_That students have developed those learning skills necessary to undertake further studies with a high degree of autonomy
T2_That students have the ability to work as members of an interdisciplinary team either as one more member, or performing management tasks in order to contribute to developing projects with pragmatism and a sense of responsibility, making commitments taking into account the available resources
The course is an introduction to Deep Learning in today's big data environments. Facilitating a practical approach and reducing the usual prerequisites (mathematical and statistical foundation), the course takes a tour of the explosive evolution of this artificial intelligence technique starting with the foundation in neural networks, to later introduce the concepts of MLP (Multilayer Perceptrons), CNN (Convolutional neural networks), RNN (Recurrent neural networks) among others. With an approach that combines practice (and theory) it aims to be a stimulus for the student to explore some of the methods found in many artificial intelligence solutions in our environment (chatGPT, DALLE2, autonomous vehicles, etc)
U1: Introduction to DL
Introduction to neural networks
Linear regression and optimization
Activation and backpropagation functions
Loss functions
U2: DL characteristics
Deep neural networks
Regularization techniques (L1/L2, abandonment)
Optimization algorithms (SGD, Adam, etc.)
U3: DL applied to Vision
Convolutional Neural Networks (CNN)
Pooling layers
Image classification using CNN
U4: DL applied to streaming
Recurrent Neural Networks (RNNs)
Short-term memory (LSTM) networks.
Sequence classification using RNN
U5: Transfer Learning
Transfer learning
Fine-tuning pre-trained models
Neural style transfer
U6: Generative Models
Introduction to autogenerative models
Autoencoders and Variational autoencoders (VAEs)
GAN and Diffusion Models
In order to gather evidence of the achievement of the expected learning outcomes, the following evaluative activities are carried out:
PLab = Laboratory practices [Related to all skills]
The practices will allow the student to practice concepts described in theory
Project = Deep Learning Project Presentation [Related to all skills]
Students will present a group project describing each and every stage they have developed. The code, the working document and the presentation will be delivered
Evidence of learning outcomes: All
ExiP = Exercises and class participation [Related to all skills]
Students and / or groups involved in problem solving
Exam = Examination [Related to all competencies]
Evidence of learning outcomes: All
Observations: All activities are compulsory except ExiP (Exercises and Class Participation)
The one-to-one activities presuppose the student's commitment to carry them out individually and without any collaboration with other people. All activities in which the student does not comply with this commitment to individuality will be considered suspended (grade 0), regardless of their role (sender or receiver) and without this excluding the possible application of other sanctions in accordance. with the current Disciplinary Regime.
Likewise, the activities to be carried out in groups presuppose the commitment on the part of the students who make it up to carry them out within the group and without any kind of collaboration with other groups or people who are alien (group individuality). All activities in which the group has not respected this commitment regardless of its role (sender or receiver) and without this excluding the possible application of other sanctions in accordance with the current Disciplinary Regime will be considered suspended (rating 0).
In the case of activities that can be done in groups, when in any of them the commitment of group individuality is not respected and / or fraudulent means are used in its accomplishment, the qualification of the activity will be, for all members of the group, of 0 points (Activity Note = 0) and without this excluding the possible application of other sanctions in accordance with the current Disciplinary Regime.
Any compulsory activity not delivered will be considered zero points
It is optional for teachers to accept or not deliveries outside the deadlines indicated. In the event that these late deliveries are accepted, it is up to the teacher to decide whether to apply a penalty and the amount of this.
Evaluation system
The final grade is the weighted sum of the grades for the learning activities:
Q = 0.25 PLab + 0.40 Proj + 0.10 ExiP + 0.25 Exam
Remarks on Recovery
The part of the Project (Proj) and Exam is indeed recoverable. The rest of the parts are not recoverable. For students who attend the recovery of the project, their grade (Proj) will be the one obtained in this test and their final grade (Q) will be calculated with the formulas detailed above and in no case will it be higher than 7.
Deep learning (Adaptive Computation and Machine Learning series) by Ian Goodfellow, Yoshua Bengio, and Aaron Courville (The MIT Press 2016)
Deep Learning with PyTorch: Build, Train, and Tune Neural Networks Using Python Tools by Eli Stevens, and Luca Antiga (Manning Publications 2020)