Publication in the Diário da República: Despacho n.º 8956/2023 - 31/08/2023
8 ECTS; 1º Ano, Anual, 28,0 T + 28,0 TP + 4,0 S , Cód. 393211.
Lecturer
- Rolando Lúcio Germano Miragaia (1)(2)
(1) Docente Responsável
(2) Docente que lecciona
Prerequisites
Objectives
After completing the curricular unit, students should be able to understand and apply the taught deep learning methods, namely:
C1 - Know and apply the main Python programming concepts to deep learning problems;
C2 - Understand the operation and be able to train a neural network;
C3 - Know and apply regularization and optimization methods;
C4 - Understand the operation and be able to train a convolutional network;
C5 - Understand the concept and be able to train a recurrent neural network;
Program
1. Programming in Python
1.1. Python
1.2. Machine learning platforms
1.3. TensorFlow
1.4. Keras
2. Neural networks
2.1. Structure of an artificial neuron
2.2. The perceptron
2.3. Multi-layer perceptron networks
2.4. Activation functions
2.5. Stochastic gradient descent
2.6. Backpropagation algorithm
3. Regularization and optimization
3.1. L1 and L2 regularization
3.2. Dropout
3.3. Batch
3.4. AdaGrad
3.5. Adam
4. Convolutional neural networks
4.1. Architecture of a convolutional neural network
4.1.1. Convolutional layers
4.1.2. Pooling layers
4.1.3. Fully connected layers
4.2. Training a convolutional neural network
4.3. Transfer learning
4.4. Image processing applications
5. Recurrent neural networks
5.1. Architectures:
5.1.1. RNN
5.1.2. Stacked Long-Short Term Memory (LSTM)
5.1.3. Gated recurrent units (GRU) networks
5.2. Applications
Evaluation Methodology
Project - carried out in groups with a report in article format and presentation
Oral Discussion - oral discussion grade (from 0 to 100%) to be multiplied by the project grade
Bibliography
- Chollet, F. (2021). Deep Learning with Python. --: Manning Publications
Teaching Method
Presencial/remote
- Lectures: Presentation, discussion and illustration of the syllabus
- Theoretical-practical classes: Discussion and exemplification of the syllabus and resolution of practical worksheets that address the concepts learned
Software used in class
Python
Google colab
