| Statuss(Aktīvs) | Izdruka | Arhīvs(0) | Studiju plāns Vecais plāns | Kursu katalogs | Vēsture |
| Course title | Machine Learning Basics |
| Course code | DatZ4012 |
| Credit points (ECTS) | 3 |
| Total Hours in Course | 81 |
| Number of hours for lectures | 8 |
| Number of hours for seminars and practical classes | 24 |
| Independent study hours | 49 |
| Date of course confirmation | 06/09/2022 |
| Responsible Unit | Institute of Computer Systems and Data Science |
| Course developers | |
| Dr. sc. ing., asoc. prof. Aleksejs Zacepins Dr. sc. ing., asoc. prof. Vitālijs Komašilovs |
|
| Prior knowledge | |
| DatZ1010, Introduction to Programming II DatZ3019, Algorithms and Structures Mate1038, Mathematics III |
|
| Course abstract | |
| Aim of the course is to teach students the basics about the machine learning and modern techniques in this field by researching various regression models, image classification and object detection problems, natural language processing and reinforcement learning approaches. Students learn how to work with the machine learning framework Tensorflow, develop new machine learning models as well as use existing models. Students learn the cloud platform for model running and learning. | |
| Learning outcomes and their assessment | |
| 1.To know about machine learning technologies and neural networks – practical assignments
2.To have skills to use cloud platform for machine learning task solving, skills to develop and use machine learning models – practical assignments 3.To have competence to individually search for solution for the defined task and ability to explain chosen approach – individual work and presentation |
|
| Course Content(Calendar) | |
| Lectures:
1.Machine learning (ML) basics. Main ML approaches. Data preparation and model training process. Introduction to Google Colaboratory. Linear models: training and evaluation. – 1h 2.Introduction to neural networks. Artificial neutral networks (ANN). ANN types and properties. Neuron activation functions. ANN training. – 1h 3.Convolution neural networks (CNN). CNN layers and typical architectures. CNN performance boosting approaches. Pre-trained CNN models and their application. – 1h 4.Object detection models and their use. Object annotations. Architectures of the object detection models. – 1h 5.Sequential data. Models for sequence processing. Recurrent neural networks (RNN). Training the RNN models. Long short-term memory (LSTM). – 1h 6.Natural language processing. Written text pre-processing. Word embedding. Text classification and generation models. – 1h 7.Reinforcement learning. Agent, environment and reward. Agent modelling. Q-learning algorithm. Deep Q-learning and its performance boosting approaches. – 1h 8.Presentation of the students’ individual work and discussion. – 1h Practical seminars: 1.Introduction to Python and Colab. Tensorflow basics. Model evaluation and training. Linear regression implementation. Logistic regression implementation. Feature crosses. – 6h 2.Artificial neural networks (ANN). Neuron and their links implementation. Training loop implementation. ANN evaluation. Convolution neural networks (CNN). Convolution layer implementation. Convolution layer visualization. – 6h 3.Natural language processing. Text data set preparation. Word embedding implementation. Word similarity model implementation and training. Text generation model implementation and training. – 6h 4.Reinforcement learning (RL). Discrete environment example. Q-learning implementation and agent training. Continuous environment example. Deep Q-learning model implementation and agent training. – 6h |
|
| Requirements for awarding credit points | |
| All practical tasks should be completed. Individual work should be completed and presented. | |
| Description of the organization and tasks of students’ independent work | |
| During the course, students should prepare and present an individual work, which implies development of the machine learning model for detection of at least 3 chosen object classes. Students should read additional literature and web resources on the course subject. | |
| Criteria for Evaluating Learning Outcomes | |
| Test with grade. Mark for this course is an accumulated mark, which includes attendance of the lectures and practical assignments, completion of the practical tasks and presentation of the individual work. | |
| Compulsory reading | |
| 1.Haykin S. S. et al. Neural networks and learning machines. Vol. 3. Upper Saddle River: Pearson, 2009. 2.Graupe D. Deep learning neural networks: design and case studies. World Scientific Publishing Company. New Jersey: World Scientific, 2016. 263 p. | |
| Further reading | |
| 1.Alpaydin E. Machine learning: the new AI. Cambridge, MA: MIT Press, 2016. | |
| Periodicals and other sources | |
| 1. Colaboratory. Pieejams: https://colab.research.google.com
2.The Jupyter Notebook. Pieejams: http://jupyter.org/ 3.TensorFlow. Pieejams: https://www.tensorflow.org/ |
|
| Notes | |
| Computer Control and Computer Science | |