EventoWeb

Advanced machine learning (ML) solutions based on deep learning (DL) have shown impressive performance on challenging, practically relevant problems in many different domains and are also present in increasingly many aspects of our lives, such as image analysis (face recognition, quality control); speech technologies (virtual assistants, emotion recognition); time series analysis (investment modeling, sensor data fusion); personalized healthcare (diagnosis, drug design); personalized advertising; or autonomous vehicles. Deep neural networks have thus become a critical component of computing in general and are an important element of many job profiles in computer and data science. 

In this module we will provide a thorough introduction to deep learning, starting by covering the fundamentals (main concepts, basic building blocks of neural networks, backpropagation, gradient descent, optimization) that will lay the way for the deep dive into the details of modern deep neural architectures, such as different variants of convolutional neural networks, recurrent neural networks, graph neural networks, and transformers. We will investigate these models and algorithms, analyze their advantages, limitations and discuss the challenges. 
 

List of topics covered:

• Introduction to DL (e.g., current applications and success, fundamental principles, deep feedforward networks, gradient-based learning, backpropagation, optimization, regularization)
• Convolutional neural networks (e.g., VGG, ResNet)
• Sequence modeling (e.g., RNN, LSTM, attention mechanism and transformers)
• Graph neural networks (e.g, GCN)
• Generative models
• Selected topics (e.g., deep RL, explainable AI, meta learning)

Accompanying Labs

To apprehend the proper application of the deep learning methods in practice, we will discuss real-world applications from several different domains (e.g., computer vision, natural language processing, healthcare) and provide a large set of practical examples on real-world datasets using state-of-the-art tools and frameworks. In a real-world practical project the participants can apply and deepen the acquired knowledge and skills. 

• Lecture slides
• Lab descriptions

• Simon J.D. Prince: "Understanding Deep Learning", MIT Press, 2023 https://udlbook.github.io/udlbook/  
• I. Goodfellow, Y. Bengio, and A. Courville: “Deep Learning”, MIT press, 2016, https://www.deeplearningbook.org/
• S. Raschka, Y. Liu, V. Mirjalili: “Machine Learning with PyTorch and Scikit-Learn: Develop machine learning and deep learning models with Python”, Packt Publishing, 2022.  https://sebastianraschka.com/blog/2022/ml-pytorch-book.html
• Scientific literature as introduced in the course

• General introduction to machine learning and neural networks, such as successful completion of MLDM1
• Python

() German (X) English

( ) Yes (X) No

Type 3a

The module description is part of the legal basis in addition to the general academic regulations. It is binding. During the first week of the semester a written and communicated supplement can specify the module description in more detail.