Workshop Deep Learning

Google, Facebook and the University of Groningen, we're all deep learning and we'd like to share the fun! If you have some experience with Python, this course introduces you to designing deep learning models with Keras. You'll get to play with Feed Forward Neural Networks in an environment that we provide during the exercises. Even though deep learning relies on mathematics, we avoid it. If you know your maths, you'll be able to fill in these blanks and if not you get to focus on how to use deep learning.

Contents

We will discuss:

• train / test / validation data set splits
• learning curves
• learning rates
• weight initialization
• gradient decent optimization methods for deep learning
• regularization
• overfitting
• convolutional layers
• pre-trained networks for image classification

Furthermore:

• the course includes lectures and practicals
• you can work on a RUG computer (or on your laptop)
• feel free to watch our deep learning talk: https://streaming3.service.rug.nl/p2gplayer/Player.aspx?id=g4tW
• we might add and exercise on recurrent neural networks