|
Lecturer(s)
|
-
Paleček Karel, Ing. Ph.D.
|
|
Course content
|
Lecture and lab topics: 1) Introduction, linear classification 2) Logistic regression and SVM 3) Multi-layer networks and back-propagation 4) Convolutional neural networks 5) Training neural networks: initialization, non-linearities, regularization 6) Training neural networks: data normalization and augmentation, verification 7) Recurrent neural networks 8) Text generation and classification 9) Object localization and detection 10) Object segmentation and image enhancement 11) Visualizaton and analysis of CNNs, style transfer 12) Deep generative models, unsupervised learning 13) Overview of neural networks libraries and frameworks 14) Introduction to reinforcement learning
|
|
Learning activities and teaching methods
|
Monological explanation (lecture, presentation,briefing), Self-study (text study, reading, problematic tasks, practical tasks, experiments, research, written assignments)
- Class attendance
- 56 hours per semester
- Home preparation for classes
- 64 hours per semester
- Preparation for exam
- 30 hours per semester
|
|
Learning outcomes
|
The class will focus on deep learning with neural newtorks (NN), currently a hot topic in machine learning and artificial intelligence. Students will learn both theoretic fundamentals of NNs as well as their real world applications such as object detection and recognition in computer vision or automatic language translation.
Theoretic piece of knowledge and practical skills from requered areas.
|
|
Prerequisites
|
The prerequisite is the completion of the subject Introduction to Machine Learning (USU).
|
|
Assessment methods and criteria
|
Combined examination
Students are required to complete all mandatory assignments.
|
|
Recommended literature
|
-
https://www.deeplearning.ai/.
-
Bishop, C. Pattern Recognition and Machine Learning. 2006. ISBN 13: 978-038731073.
-
Goodfellow, I., Bengio, Y., Courville, A. Deep learning. MIT Press, 2016.
-
Karpathy, A., Johnson, J., Li, F. Convolutional neural neworks for visual recognition.
|