Representation Learning and Unsupervised Learning with Deep Neural Networks
Although Deep Learning is known to be very successful in Supervised Learning it is also perfectly applicable to Unsupervised Learning. You can do a lot of great things with unlabelled data and Deep Neural Networks.
Two days of hands-on training containing:
- Get acquainted with the topic and build a solid understanding of this very relevant field.
- Experiment with the main Deep Learning building blocks for Representation Learning and Unsupervised Learning.
- Explore the notion of latent spaces and embeddings. Here we will have a look how you can map almost any data to lists of numbers. And we will also find how that these numbers can be used for computing distances, similarities and differences.
- Learn about Generative Adversarial Networks. These allow you to create new data from your data-sets.
- Understand and apply Autoencoders and Variational Autoencoders. How to apply Deep Learning to clustering? How would you build a recommender system using Deep Learning?
- Investigate the powerful technology called Triplet-Loss. It allows you to solve identification tasks with a very small amount of data and a high accuracy.