30th Deep Learning Meetup in Vienna: Fake News

Dear Deep Learners,

Fake news is a big topic these days and most likely you have also heard about Deep Fakes. At our next meetup on October 29th we will have a special talk about Deep Fakes and fake news. In addition, we will hear about Anomaly Detection with GANs.

Talk 1:
Fake News. From Shallow to Deep. How to create, detect and fight it.
Alexander Schindler, Research Scientist, Austrian Institute of Technology (AIT)

Deep Fakes impressively demonstrated the potential of deep learning for media production. This potential also raises concerns and fears that this technology could be misused to create and disseminate misinformation. The few well engineered examples available underline these concerns and light-headed projects such as "Deep Nude" further incite them. But how do these technologies work? Are there any flaws that can be used to automatically detect fake media and are Deep Fakes the only threats or are there more options to fake news? This talk will provide an overview of approaches to fake news and how to detect those.

Alexander Schindler is a research scientist at Austrian Institute of Technology (AIT) and member of the Music Information Retrieval Lab at the TU Wien. His research fields are audio, audio-visual and multi/cross-modal analysis of media.

Talk 2:
Anomaly Detection with GANs
Thomas Schlegl, Machine learning Engineer and Data Scientist at contextflow

The detection and quantification of anomalous patterns in medical images that correlate with disease status is very important during diagnosis and enables monitoring of disease progression or treatment response. This talk presents an anomaly detection technique that is based on Generative Adversarial Networks (GANs) and its application in medical image analysis. A GAN learns to generate data that follows the distribution seen during training. The presented anomaly detection framework trains a GAN on normal images to create a rich generative model of healthy local anatomical appearance and allows for quantitative assessment of new images using the learned generative model.

Thomas Schlegl is a machine learning engineer and data scientist at contextflow GmbH - a company building an image search engine for radiologists. His research interests focus on medical image analysis, biostatistics, and multimodal learning of deep representations based on visual and textual clinical data.

Hot Topics:
As usual we will feature latest news and hot topics about Deep Learning. Moreover, Elisabeth Fink will present the best talks from the O'Reilly AI conference in San Jose, CA in September.

This meetup is kindly hosted by A1 Telekom.

Looking forward to welcoming you,
Tom, Alex, Jan, René