Study Guide

Objective

After completing the course, the student can:
- Can define the main concepts, values and drivers for deep learning
- Can describe how machine learning and AI solutions can be developed with deep learning and neural networks
- Use tools when creating the solutions

Content

Deep Learning
Neural Networks
Natural Language Processing
Pattern Recognition
Computer Vision
Practical work

Materials

Course materials are prepared by the lecturer from various sources including books, online material, etc.

Recommended books to study in this course are:
-- Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow: Concepts, Tools, and Techniques to Build Intelligent Systems 2nd Edition by Aurélien Géron, OREILLY, 2021
-- Deep Learning with Python, François Chollet

Exam schedules

Exam will be in the first week of December (at the same date as we have the regular lectures).

International connections

The course includes 12 theory sessions and personal practice tasks.

The lectures cover the main theories, techniques, and algorithms in basics of deep learning, starting with fundamental concepts such as neural networks, optimization and regularization techniques, modelling, and fine tuning. Through the course, there will be more practical applications of deep learning such as CNNs, RNNs, Computer Vision, and NLP.

Completion alternatives

The exercises are mainly performed using Jupyter Notebook. Student will use TensorFlow and/or PyTorch. Strong python programming is need to complete the exercises.

Student workload

12 sessions (2.9-29.11.24 ) each 3 hours (1h lecture, 2h practice)+ Exam

Contact hours:
- Course start-up (week 36): 2h
- Weeks 37 - 48: Theory & practice (3h/week): 12 x 3h = 36h
- Week 49: Exam: 2h
- In addition, about 5 support and inquiry hours (biweekly): 5x 1h = 5h

Total contact hours: 45 hours
Independent study and homework: about 90 h

Content scheduling

The course will cover the following concepts:
-- Introduction to Deep Learning
-- Tensors and tensor operations
-- Multi layer Perceptron
-- Gradient based Optimization
-- Back-propagation
-- Loss Functions
-- Activation Functions
-- Convolutional Neural Networks
-- Recurrent Neural Networks
-- DNN Architectures
-- Hyperparameter Fine-Tuning
-- Transfer Learning
Practical Aspects of Deep Learning will be also covered during the course and exercise.

Evaluation scale

H-5

Assessment methods and criteria

After the course, students should understand the main principles of deep learning and steps needed for applying them in real applications. The student especially learns the core concepts of deep neural networks, gradient descent, model evaluation, overfitting, and underfitting and is able to find a suitable balance between these extremes in a given problem at hand.

This course comprises 100 study points including:
-- 24 points (1+1p each contact class: Lecture and Practical Session)
-- 36 points for exercises
-- 40 points for the exam

-Participation and exercise (50% of total to pass): Students must achieve at least 50% of the points to pass the course. Participation is

- Exam (50% of total points to pass): Students must achieve at least 50% of the points (20 points) in order to pass the course.

The course is graded on a scale of 0-5.

Grading will be according to the total points collected by the student during the course as well as the exam.
1: 50% (minimum to pass the course)
2: 60-70%
3: 70-80%
4: 80-90%
5: 90- 100%