Introduction to Deep Learning

Welcome to the web page of the class Introduction to Deep Learning at the Technical University of Košice, a course that is taught in the winter semester in the first year of MSc. studies for students of Intelligent Systems. The course is a continuation of the course Neural networks. This web page provides all necessary information and materials for the course.

Your teachers

• prof. Peter Sinčák
• Ján Magyar, MSc.
• Andrij David

Grading

To successfully pass this course, you have to meet the following requirements:

1. attendance at lectures and labs (at most 3 absences at each)
2. hand in the semestral project (see lower)
3. get at least 21 points during the semester
4. pass the exam (get at least 31 points)

Throughout the semester you can get 40 points in the following distribution:

• semestral project - 30 points
• homework and class activities - 10 points

Course plan

Due to the ongoing COVID-19 pandemic, teaching will take place in a hybrid system with both online and in attendance classes. All lectures will be online, while labs will be online and also in attandance (see table lower). For in attendance classes, the study group will be split into two smaller halves (with no more than 12 students), which will attend classes alternating. In attendance classes will take place at 7.30 on Mondays in V4_V147. Additional online consultations are offered Fridays at 9.00 on Microsoft Teams.

Please check the course's Microsoft Teams group for the newest information.

	Lecture	Lab
Week 1 21. 9. - 27. 9.	TBA	introduction to the course [online]
Week 2 28. 9. - 4. 10.	TBA	review of ANNs, theory of convolution [1st group]
Week 3 5. 10. - 11. 10.	TBA	convolution - implementation [online]
Week 4 12. 10. - 18. 10.	TBA	CNNs in keras, visualizing CNNs [online]
Week 5 19. 10. - 25. 10.	TBA	DL training cookbook [online]
Week 6 26. 10. - 1. 11.	TBA	project report [online]
Week 7 2. 11. - 8. 11.	TBA	text processing and recurrent neural nets [online - in English]
Week 8 9. 11. - 15. 11.	TBA	DL for tabular data [online - in English]
Week 9 16. 11. - 22. 11.	TBA	autoencoders [online]
Week 10 23. 11. - 29. 11.	TBA	Generative Adversarial Nets [online]
Week 11 30. 11. - 6. 12.	TBA	deep reinforcement learning [online]
Week 12 7. 12. - 13. 12.	TBA	handing in assignments [online presentation]
Week 13 14. 12. - 20. 12.	—	handing in assignments [online presentation]

Assignment

During the semester, each student must participate in the completion of an assignment. Assignments are done by teams of three or four students, within the team, students will be competing against each other in training the most accurate network (every student will prepare his/her own solution). Each team project must contain the following:

• a specified research goal (per team)
• an overview of state-of-the-art solutions (per team)
• a trained DL model (per student)
• documentation (per student)
• a research paper presenting the results of the team project (per team - the best papers can be later published in journals).

For further information on expected outcome and grading, please check this page.

Topics:

1. image classification
2. segmentation
3. superresolution
4. image inpainting
5. image coloring
6. deep reinforcement learning
7. time series prediction
8. image noise reduction

Further sources

• Deep Learning Simplified
• Maths of Intelligence
• Convolutional Neural Networks
• Intro to Deep Learning
• Hluboké neuronové sítě