Module also offered within study programmes:
General information:
Name:
Machine Learning in Networking
Course of study:
2019/2020
Code:
ITEI-2-305-s
Faculty of:
Computer Science, Electronics and Telecommunications
Study level:
Second-cycle studies
Specialty:
-
Field of study:
ICT studies
Semester:
3
Profile of education:
Academic (A)
Lecture language:
English
Form and type of study:
Full-time studies
Course homepage:
 
Responsible teacher:
dr inż. Bułat Jarosław (kwant@agh.edu.pl)
Module summary

Student will be able to design a computer system based on Machine Learning algorithms. It will make use of contemporary software libraries as well as methods and techniques. Student will understand various concepts of ML like: optimization algorithms, types of ML algorithms, capabilities and limitation of these methods.

Description of learning outcomes for module
MLO code Student after module completion has the knowledge/ knows how to/is able to Connections with FLO Method of learning outcomes verification (form of completion)
Social competence: is able to
M_K001 Student can critically and creatively approach a network or system design problem, find the potential of using novel methods characteristic for machine learning related to it, pose it in a clear way, and analyze it on her/his own while discussing the topic in a group. TEI2A_K03, TEI2A_K01 Involvement in teamwork
Skills: he can
M_U001 Student can design a system using ML methods according to adopted assumptions or constraints. TEI2A_U05, TEI2A_U02, TEI2A_U01 Execution of laboratory classes
M_U002 Student can use contemporary computer library and tools for implementing machine learning algorithms. TEI2A_U06, TEI2A_U04 Execution of laboratory classes
M_U003 Student can choose the appropriate algorithm of machine learning for given class of problem. TEI2A_U05, TEI2A_U06, TEI2A_U04 Execution of laboratory classes
Knowledge: he knows and understands
M_W001 Student gets extended and integrated knowledge (notions and methods) in the area of application of various concepts of Machine Learning in ICT. TEI2A_W01 Test results
M_W002 Student gets extended and integrated knowledge (notions and methods) in the area of application of various concepts of optimization algorithm in Machine Learning context. TEI2A_W01 Test results
M_W003 Student gets extended and integrated knowledge (notions and methods) in the area of application of various concepts of neural networks in ICT. TEI2A_W01 Test results
Number of hours for each form of classes:
Sum (hours)
Lecture
Audit. classes
Lab. classes
Project classes
Conv. seminar
Seminar classes
Pract. classes
Zaj. terenowe
Zaj. warsztatowe
Prace kontr. przejść.
Lektorat
60 15 0 30 0 0 15 0 0 0 0 0
FLO matrix in relation to forms of classes
MLO code Student after module completion has the knowledge/ knows how to/is able to Form of classes
Lecture
Audit. classes
Lab. classes
Project classes
Conv. seminar
Seminar classes
Pract. classes
Zaj. terenowe
Zaj. warsztatowe
Prace kontr. przejść.
Lektorat
Social competence
M_K001 Student can critically and creatively approach a network or system design problem, find the potential of using novel methods characteristic for machine learning related to it, pose it in a clear way, and analyze it on her/his own while discussing the topic in a group. - - + - - + - - - - -
Skills
M_U001 Student can design a system using ML methods according to adopted assumptions or constraints. - - + - - + - - - - -
M_U002 Student can use contemporary computer library and tools for implementing machine learning algorithms. - - + - - + - - - - -
M_U003 Student can choose the appropriate algorithm of machine learning for given class of problem. - - + - - + - - - - -
Knowledge
M_W001 Student gets extended and integrated knowledge (notions and methods) in the area of application of various concepts of Machine Learning in ICT. + - - - - - - - - - -
M_W002 Student gets extended and integrated knowledge (notions and methods) in the area of application of various concepts of optimization algorithm in Machine Learning context. + - - - - - - - - - -
M_W003 Student gets extended and integrated knowledge (notions and methods) in the area of application of various concepts of neural networks in ICT. + - - - - - - - - - -
Student workload (ECTS credits balance)
Student activity form Student workload
Summary student workload 100 h
Module ECTS credits 4 ECTS
Udział w zajęciach dydaktycznych/praktyka 60 h
Preparation for classes 30 h
przygotowanie projektu, prezentacji, pracy pisemnej, sprawozdania 10 h
Module content
Lectures (15h):

  1. Introduction, terminology, problem statement, supervision learning, unsupervised learning, prediction vs classification, description of concepts: training set, learning algorithms, hypothesis (model, parameters), cost function, linear regression (one variable, multiple variables), different models (linear, polynomial, …), logistic regression.
  2. Neural Network representation, initialization, backpropagation algorithm (symmetry breaking problem), introduction to minimization algorithms and cost functions.
  3. Advanced method of minimization algorithms.
  4. Selected neural network architecture: deep neural networks, convolutional, recurrent.
  5. Selected neural network architecture: resnet, attention, autoencoder.
  6. Reinforcement learning.

Laboratory classes (30h):

  1. Introduction, Python in ML, loading data, preparation (cleaning, formatting), introduction to tensorflow in Python (in colab environment https://colab.research.google.com/).
  2. Basic problems in ML with examples in playground (https://playground.tensorflow.org)
  3. Using tensorflow to linear regression and logistic regression – introduction to supervised learning.
  4. Visualization, logging, checkpointing in tensorflow with link to “tensorboard”
  5. Using fully connected NN for pattern recognition for MNIST classification.
  6. Using convolution for image recognition for traffic light recognition.
  7. Recurrent NN for sequences with examples in text processing. As an example prediction of missing word in text will be presented.
  8. Bias-Variance trade-off – how to test your classificator and model. How to detect and prevent low variance (high bias) and high variance (overfitting). Methods for avoiding overfitting (dropout, regularization).
  9. Transfer learning in examples based on tf-hub. Modification of architecture of an existing model without re-training or with limited re-training. Reusability trained model.
  10. Embedding and advanced visualization.
  11. Unsupervised learning – autoencoder, variational autoencoder. Problem of dimensionality reduction on an example of image compression and representation learning.
  12. Introduction to reinforcement learning as an example of machine learning driven control. Implementing pong bot.
  13. Developing ML application solving practical problem – data preparation, data cleaning, model designing.
  14. Developing ML application solving practical problem – model training, evaluation.
  15. Evaluation.

Seminar classes (15h):

Seminar classes will be a practical introduction to laboratories and cover topics of laboratory classes.

Additional information
Teaching methods and techniques:
  • Lectures: The content presented during the lecture is provided in the form of a multimedia presentation in combination with a classical lecture panel.
  • Laboratory classes: During the laboratory classes, students in teams solve the practical problem, choosing the right tools and methods.
  • Seminar classes: Seminar classes are based on a multimedia and oral introduction to the topic and practical implementation of tasks provided by lecturer. An important elements of this activities are: team working, discussion and presentation of results.
Warunki i sposób zaliczenia poszczególnych form zajęć, w tym zasady zaliczeń poprawkowych, a także warunki dopuszczenia do egzaminu:

Seminar
It is necessary to prepare application solving practical problem with the help of ML algorithm. Application will be evaluated at 50 points. The assessment criterion will be the quality of the results obtained and the short report.

Laboratory
Two test (30 minutes) will be scheduled during laboratory. Each test will be evaluated at 50 points. Grade from laboratory will be calculated according to the Study Regulation from sum of test grades.

Participation rules in classes:
  • Lectures:
    – Attendance is mandatory: No
    – Participation rules in classes: Students participate in the classes learn topics according to the syllabus. Students should ask questions and explain doubts. Audiovisual recording of the lecture requires the teacher's consent.
  • Laboratory classes:
    – Attendance is mandatory: Yes
    – Participation rules in classes: Students carry out laboratory exercises in accordance with materials provided by the teacher. The student is obliged to prepare for the subject of the exercise, which can be verified in an oral or written test. Completion of classes takes place on the basis of presenting a solution to the problem.
  • Seminar classes:
    – Attendance is mandatory: Yes
    – Participation rules in classes: Students carry out exercises introduced by the teacher and participate in the discussion on this topic. Students work in groups (2-3 person)
Method of calculating the final grade:

The Seminar and Laboratory grades will be taken into account to calculate final grade. Grades from Seminar and Laboratory have to be positive. Final grade will be determined on the basis of the weighted average of grades from Seminar and Laboratory, then the thresholds defined in the Study Regulation will be applied.

Sposób i tryb wyrównywania zaległości powstałych wskutek nieobecności studenta na zajęciach:

In case of missing evaluation test, an additional term will be scheduled at the end of the semester.

Prerequisites and additional requirements:

- knowledge of linear algebra
- programming skills in Python
- basic knowledge of optimization methods
- basic knowledge of data mining

Recommended literature and teaching resources:
  1. E. Sanchez, G. Squillero, A. Tonda, Industrial Applications of Evolutionary Algorithms, Springer 2012.
  2. J. Watt, R. Borhani, A. K. Katsaggelos, Machine Learning Refined, Cambridge University Press 2016.
  3. L. Torgo, Data Mining with R, CRC Press 2017.
  4. Andrew NG, Machine learning yearning – technical strategy for AI Engineers, in the Era of Deep Learning, draft – deeplearning.ai
Scientific publications of module course instructors related to the topic of the module:
  1. K. Rusek, P. Chołda, Message-Passing Neural Networks Learn Little’s Law, IEEE Communications Letters, February 2019, volume 23, is. 2.
  2. K. Rusek, J. Suárez-Varela, A. Mestres, P. Barlet-Ros, and A. Cabellos-Aparicio. 2019. Unveiling the potential of Graph Neural Networks for network modeling and optimization in SDN. In Proceedings of the 2019 ACM Symposium on SDN Research (SOSR ’19)
  3. K. Rusek, P. Guzik, Two-stage neural network regression of eye location in face images Multimedia Tools and Applications, 1-14
  4. L. Janowski, P. Romaniak, Z. Papir Content driven QoE assessment for video frame rate and frame resolution reduction Multimedia tools and applications 61 (3), 769-786
  5. L. Janowski, P. Kozłowski, R. Baran, P. Romaniak, A. Glowacz, T. Rusc, Quality Assessment for a Visual and Automatic License Plate Recognition, Multimedia Tools and Applications, January 2014, Volume 68, Issue 1, pp. 23–40.
  6. L. Janowski, M. Duplaga, K. Suwada, Using MPEG-7 Descriptors and Scoring Model to Automatically Recognize Medical Events in the Record of Bronchoscopy Video, Information Technologies in Biomedicine, 2010, pp. 547–558.
  7. L. Janowski, M. Pinson The Accuracy of Subjects in a Quality Experiment: A Theoretical Subject Model Multimedia, IEEE Transactions on 17 (12), 2210-2224
  8. L. Janowski and Z. Papir, “Modeling subjective tests of quality of experience with a Generalized Linear Model,” 2009 International Workshop on Quality of Multimedia Experience, San Diego, CA, 2009, pp. 35-40.
  9. J. Bulat et al. “Data processing tasks in wireless GI endoscopy: Image-based capsule localization & navigation and video compression.” Engineering in Medicine and Biology Society, 2007. EMBS 2007. 29th Annual International Conference of the IEEE. IEEE, 2007.
  10. J. Bułat, and A. Głowacz. “Vision-based navigation assistance for visually impaired individuals using general purpose mobile devices.” Signals and Electronic Systems (ICSES), 2016 International Conference on. IEEE, 2016.
Additional information:

The earlier lecture „Data mining” is a good introduction to this lecture.

Classes are conducted using innovative teaching methods developed during 2017-2019 in the POWR.03.04.00-00-D002/16 project, carried out by the Faculty of Computer Science, Electronics and Telecommunications under the Smart Growth Operational Programme 2014-2020.

The staff has improved communications skills, which have been developed during English language trainings in the POWR.03.04.00-00-D002/16 project, carried out by the Faculty of Computer Science, Electronics and Telecommunications under the Smart Growth Operational Programme 2014-2020.