Educational guide
IDENTIFYING DATA 2023_24
Subject DEEP LEARNING FOUNDATIONS Code 01747015
Study programme
1747 - Máster Universitario de Investigación en Ciberseguridad
Descriptors Credit. Type Year Period
3 Optional First First
Language
Ingles
Prerequisites
Department ING.ELECTR.DE SIST. Y AUTOMATI
Coordinador
FIDALGO FERNANDEZ , EDUARDO
E-mail efidf@unileon.es
vgonc@unileon.es
fjaÑm@unileon.es
Lecturers
GONZÁLEZ CASTRO , VICTOR
FIDALGO FERNANDEZ , EDUARDO
JAÑEZ MARTINO , FRANCISCO
Web http://agora.unileon.es
General description In this course, we will study Deep Learning from scratch and how it could be applied to cybersecurity. During the course, we will explain where deep learning is inside Machine Learning and Artificial Intelligence, neural networks, how their different parts work together, and some neural network architectures. Finally, in the laboratory sessions, we will consolidate the concepts from the lectures using Python.
Tribunales de Revisión
Tribunal titular
Cargo Departamento Profesor
Presidente ING.ELECTR.DE SIST. Y AUTOMATI BLAZQUEZ QUINTANA , LUIS FELIPE
Secretario ING.ELECTR.DE SIST. Y AUTOMATI ALAIZ MORETON , HECTOR
Vocal ING.ELECTR.DE SIST. Y AUTOMATI FUERTES MARTINEZ , JUAN JOSE
Tribunal suplente
Cargo Departamento Profesor
Presidente ING.ELECTR.DE SIST. Y AUTOMATI PRADA MEDRANO , MIGUEL ANGEL
Secretario ING.ELECTR.DE SIST. Y AUTOMATI FOCES MORAN , JOSE MARIA
Vocal ING.ELECTR.DE SIST. Y AUTOMATI GARCIA RODRIGUEZ , ISAIAS

Competencies
Type A Code Competences Specific
  A18801
Type B Code Competences Transversal
  B5729
  B5730
  B5731
  B5732
  B5740
Type C Code Competences Nuclear
  C2
  C4
  C5

Learning aims
Competences
Knowledge of the scientific method A18801
 B5729
B5730
B5732
B5740
 C5
Ability to search for relevant information and references and writing scientific articles B5729
B5730
B5731
B5732
B5740
 C2
C4
C5
Preparation of contributions for scientific conferences A18801
 B5732
 C5

Contents
Topic Sub-topic
Block I. THEORETICAL CONCEPTS Lesson 1. INTRODUCTION TO DEEP LEARNING
What is deep learning. When to use it. Advantages and disadvantages.

Lesson 2. NEURONAL NETWORKS
Basic concepts. Architectures. Monolayer neural networks. Deep neural networks

Lesson 3. DEEP LEARNING FOR IMAGE PROCESSING
Convolutional neural networks. Creation and adjustment of convolutional neural networks. Learning curves. Interpretation of models.

Lesson 4. MACHINE AND DEEP LEARNING FOR NATURAL LANGUAGE PROCESSING
NLP concepts. Classical methods. Data preprocessing. Word Embeddings. Text classification.

Lesson 5. OTHER DEEP LEARNING MODELS
Architecture categories for deep learning. Recurrent Neural Networks. Unsupervised pretrained networks.
Block II. PRACTICAL SESSIONS The practical sessions will be carried out using Python 3.X and a Deep Learning framework, such as Keras or PyTorch. An environment with these characteristics will be available in the laboratory, although it is recommended that the student install it on their own computer.
There will be at least one practical session related to each of the Topics seen in theory, where the student will solve problems in which the concepts seen will be applicable.

Planning
Methodologies  ::  Tests
  Class hours Hours outside the classroom Total hours
Practicals using information and communication technologies (ICTs) in computer rooms 14 14 28
 
Case study 0 14 14
Presentations / expositions 2 3 5
 
Lecture 14 14 28
 
 
(*)The information in the planning table is for guidance only and does not take into account the heterogeneity of the students.

Methodologies
Methodologies   ::  
  Description
Practicals using information and communication technologies (ICTs) in computer rooms Guided practicals in the laboratory programming in Python and applying the concepts seen in the Lecture.
Case study As a Course Project, the student will choose a topic from a list proposed, or the student will suggest another one that needs to be approved by the instructor. The topic must be related to a work involving cybersecurity and Deep Learning (not Machine Learning). The student will prepare a research proposal and reproduce a solution from a selected research paper.
Presentations / expositions Once both the theoretical and practical parts of the Course Project have been delivered following the indicated requirements and using the corresponding tasks in Agora, all the students will prepare a presentation they will give in front of all the students and the instructor
Lecture Theoretical sessions in the classroom using slides that will be recorded in videos. Presentations or documents corresponding to the materials of each lesson will be left in Agora. Some lessons may be accompanied by videos related to the concepts presented, some recorded by the teachers and others from internet resources that the teachers consider especially appropriate. Datacamp platform courses, or similar ones, could be used to reinforce some of the lessons taught, with some courses being optional and others being mandatory.

Personalized attention
 
Practicals using information and communication technologies (ICTs) in computer rooms
Lecture
Case study
Presentations / expositions
Description
Students can request personalized attention via email at any time during the course. Said attention will be provided via videoconference for students of the remote modality, if necessary.

Assessment
  Description Qualification
Practicals using information and communication technologies (ICTs) in computer rooms Each laboratory will contain deliverables that will be assessed, and a grade will be set to each laboratory session. 30
Lecture A small questionnaire will be done to evaluate some concepts given in the lectures (10). After the questionnaire, the students will do the second part of the evaluation, which will be a written exam (20) 30
Case study It will be evaluated several parts of the Course Project, the research proposal and the proof of concept. 20
Presentations / expositions It will be evaluated several parts of the Course Project that will be presented to the students, including the final presentation to be given to the instructor. 10
Others Some volunteer activities could be offered to the students. These will be evaluated to get additional points to the grade of the subject 10
 
Other comments and second call
  • Delayed deliveries will get a pensalisation in the grade.
  • To pass the course in continuous assessment (i.e. the first call) it will be necessary to get at least 5 out of 10 points.
  • Students can compensate grades between parts, as long as the minimum grade on a part is 3 out of 10.
  • Students who do not pass the course in continuous assessment (i.e. the first call) will be able to submit labs which were not submitted or which were not passed during the first call.
  • To pass the course in second call, it will be applied the same instructions than in the first call. 

For students of the ONLINE modality of the master's degree:

In reference to the supervision programs used (SMOWL) during the exams of the official calls of the distance modality, browsing in pages external to that of the exam itself, unless expressly indicated, may result in failure in said activity, at the discretion of the faculty.

In the event that problems arise in student identification, teachers may require additional assessment activities via videoconference. The conditions of these tests may be conditioned by connectivity, lighting, etc. It is the responsibility of the student to follow the instructions received in this regard, as well as to protect their privacy, performing the exam in an appropriate environment (isolated, with good connection, lighting ). , ....). Recommendations for students in the use of SMOWL can be found at the following link: http://bit.ly/3ZrtxVs

Sources of information
Access to Recommended Bibliography in the Catalog ULE

Basic
  • Aston Zhang, Zachary C. Lipton, Mu Li and Alexander J. Smola., Dive into Deep Learning, Cambridge Univerity Press, 2023
  • Cunha. (2022). Deep learning with Python (2a ed) - François Chollet - Manning, outubro 2021, 504 pp. Interações: Sociedade e as novas modernidades, 42, 113-115. https://doi.org/10.31211/interacoes.n42.2022.r1
  • Geron. (2017). Hands-on machine learning with Scikit-Learn and TensorFlow: concepts, tools, and techniques to build intelligent systems  (1st ed.). O’Reilly.
  • Bird. (2009). Natural language processing with Python (Klein & E. Loper, Eds.; 1st edition). O’Reilly
Complementary
  • Research papers whose references will be shared and updated on each course. 
  • Lewis Tunstall. (s. f.). Natural Language Processing with Transformers. https://www.oreilly.com/library/view/natural-language-processing/9781098136789/
  • Howard. (2020). Deep learning for Coders with fastai and PyTorch?: AI applications without a PhD  (Gugger, Ed.). O’Reilly Media Inc.
  • Huang, Hussain, A., Wang, Q.-F., & Zhang, R. (2019). Deep Learning for Natural Language Processing. En Deep Learning: Fundamentals, Theory and Applications (Vol. 2, pp. 111-138). Springer International Publishing AG. https://doi.org/10.1007/978-3-030-06073-2_5
  • Goodfellow. (2016). Deep learning  (Bengio & A. Courville, Eds.). The MIT Press.

Recommendations


 
Other comments
Knowledge of Python or other programming languages is recommended.