===
As we would like to have a head count as soon as possible to estimate how many students are interested in this lecture, please sign in for this lecture using the key learnDL if you plan to hear this lecture.
===
In recent years, deep learning network has steadily increased in popularity, mainly due to their state-of-the-art performance in image and speech recognition, text mining and related tasks. Deep neural networks attempt to automatically learn multi-level representations and features of data and are able to uncover complex underlying data structures.
The lecture aims at providing a basic theoretical and practical understanding of modern neural network approaches. We will start out by covering the necessary background on traditional artificial neural networks, backpropagation, online learning, and regularization. Then we will cover special methods used in deep learning, like drop-out and rectified linear units. We will also talk about further advanced topics like convolutional layers, recurrent neural networks, and auto-encoders.
We will also talk about practical application and open-source deep learning libraries.
Requirements:
English
Statistics Master (any) or Data Science Master
Some background in modeling, e.g., lecture on GLMs, preferably a lecture on machine learning / predictive modeling
Some background in optimization, e.g., Computational Methods I in the statistics master
Practical programming knowledge in R or Python
- Trainer/in: Niklas Klein
- Trainer/in: Janek Thomas