A tantárgyleírás hatályossága
Hatályosság kezdete:
2026. March 21.
Hatályosság vége:
—
| Subject name (Hungarian, English) |
Mélytanulás
Deep Learning
|
||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Subject code | BMEVITMMA19 | ||||||||||||
| Subject type | — | ||||||||||||
| Training Level | — | ||||||||||||
| Course types and hours (weekly/semester) |
|
||||||||||||
| Assessment type | vizsga | ||||||||||||
| Credits | 5 | ||||||||||||
| Subject coordinator |
DR. Gyires-Tóth Bálint Pál
position: egyetemi docens
contact:
gyires-toth.balint@vik.bme.hu
|
||||||||||||
| Responsible department |
Távközlési és Mesterséges Intelligencia Tanszék
|
||||||||||||
| Faculty | Villamosmérnöki és Informatikai Kar | ||||||||||||
| Subject website | — | ||||||||||||
| Primary curriculum type | — | ||||||||||||
| Direct prerequisites – Strong prerequisite | none | ||||||||||||
| Direct prerequisites – Weak prerequisite | none | ||||||||||||
| Direct prerequisites – Parallel prerequisite | none | ||||||||||||
| Direct prerequisites – Milestone prerequisite | none | ||||||||||||
| Direct prerequisites – Exclusion | none |
Objectives
Programme
Lectures:
- Background of deep learning; the principal software architecture of deep learning (e.g. Google Colab, NVIDIA Docker, Kubernetes, SLURM, TensorFlow, TensorFlow-Keras, PyTorch) and its hardware architecture; sizing and scaling.
- Forward propagation, backward propagation, loss functions, optimisation methods, activation functions, GPU implementation.
- Weight initialisation; regularisation methods in deep neural networks (e.g. early stopping, dropout, batch normalization, layer normalization, etc.).
- Fundamental hyperparameters and their effect on training (e.g. number of epochs, batch size, activation functions, optimisation algorithms and learning rate, etc.).
- Overview of deep learning framework modules; neural networks as computational graphs; training of parameters.
- Automatic differentiation in deep learning frameworks; complex connections (residual and skip connections); inspection and monitoring of data and training (e.g. using TensorBoard and WandB.ai).
- Fundamentals of convolutional neural networks; backpropagation in convolutional layers; pre-trained neural networks; transfer learning.
- Popular convolutional neural network architectures; collections of pre-trained models; binary inference modules.
- Fundamentals of recurrent neural networks; low- and high-level implementations (Python, C++ and cuDNN).
- Sequence-to-sequence (seq2seq) models; the transformer architecture for modelling sequential data; large language models.
- Self-supervised learning (SSL) in natural language processing; transformer-based NLP applications.
- Self-supervised learning in computer vision; contrastive and non-contrastive methods.
- Deep reinforcement learning; Deep Q-Learning and related software tools.
- Fundamentals of graph neural networks and their software tools.
- GPU-based containerisation and selection of project work.
- Labelling methods; consensus-based labelling.
- Defining and implementing the evaluation methodology.
- Building baseline models.
- A deep learning-based classification and/or regression task with evaluation.
- Incremental modelling.
- Evaluation and interpretation of results; integration of a user interface (UI).
Deep learning is one of the main technologies of today's data-driven artificial intelligence methods. One of the primary advantages of deep learning over other machine learning methods is that, in a single step, it learns both the representations that best describe the data and the modelling of these representations. The deep learning paradigm has achieved state-of-the-art results in numerous scientific fields - for example in computer vision, natural language processing and speech technology. Under controlled conditions, in many applications these methods are able to approach human performance, and in some cases they even surpass it.
The research and development of deep learning systems is now supported by a wide range of hardware and software architectures. Their effective use requires a solid understanding of deep learning theory and of the relevant software and hardware tools, as well as knowledge gained through hands-on experience.
The aim of the course is to present the necessary theoretical foundations and, through practical examples, to help students master and effectively use modern deep learning software tools and techniques. The course primarily uses the open-source, Python-based deep learning frameworks PyTorch and TensorFlow / Keras.
Learning outcomes
Ez a tantárgy a KKK rendeletben meghatározott, következő kompetenciák fejlesztését szolgálja:
Knowledge
No learning outcomes recorded.
Skills
No learning outcomes recorded.
Attitudes
No learning outcomes recorded.
Autonomy and responsibility
No learning outcomes recorded.
Oktatási módszertan
Lecture
and practice.
Study of the associated assigned materials (e.g. IPython notebooks, source code,
descriptions, videos).
Tanulástámogató anyagok
Online források
Russell, S., & Norvig, P. (2021). Artificial Intelligence: a modern approach, 4th US ed. URL: https://aima.cs.berkeley.edu ; ; Chollet, F., Deep Learning with Python, Third Edition, Manning Publications, 2025; https://www.manning.com/books/deep-learning-with-python-third-edition; Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep learning. MIT press.; https://www.deeplearningbook.org/
Recommended preliminary knowledge for completing the subject
Knowledge type competencies
(azon előzetes ismeretek összessége, amelyek megléte nem kötelező, de a tantárgy eredményes teljesítését nagyban elősegíti)
Basic statistics, data structures, algorithms, fundamental concepts of artificial intelligence, basic programming skills
Skill type competencies
(azon előzetes képességek és készségek összessége, amelyek megléte nem kötelező, de a tantárgy eredményes teljesítését nagyban elősegíti)
nincs
Recommended (non-compulsory) preliminary competencies
(azon ajánlott (nem kötelező) előzetesen megszerzendő kompetenciák összessége, amelyek jelentősen hozzájárulnak a tantárgy eredményes teljesítéséhez)
Basic statistics, data structures, algorithms, fundamental concepts of artificial intelligence, basic programming skills
General rules
Requirements:
During the study period, students carry out project work. Over the course of the semester, students must report on their progress. The completed solution, together with its documentation, must be submitted at the end of the semester.
A student may receive the end-of-semester signature recognising completion of the semester only if the submitted project work meets the minimum requirements.
The course concludes with an examination. The end-of-semester grade is determined on the basis of the result achieved in the examination. A prerequisite for passing the examination (a grade of at least 'pass') is meeting the specified minimum requirement, namely achieving at least 40% of the maximum attainable score. Outstanding student performance during the study period may earn bonus points that count towards the examination result.
Additional possibilities:
The project work may be made up until the end of the make-up (retake) week.
Assessment methods
In-term assessments
No detailed assessments provided.
Weight of in-term assessments
No weights provided.
Exam-period assessments
No detailed assessments provided.
Weight of exam elements
No weights provided.
Grade calculation
No grade thresholds provided.
Attendance requirements
No attendance requirements provided.
Rules for retake and resubmission
Not provided.
Short description
Not provided.
Detailed description
Not provided.
Recommended courses
-
Workload to complete the subject
No workload breakdown provided.
Validity of subject requirements
Requirements valid from:
—
Requirements valid until:
—
Curriculum placement
No curriculum placements recorded for this subject version.