Berlin Bioinformatics Hackathon

Feb 8, 2018

In this lecture we discussed the newly released capsule network by Geoffrey Hinton and his co-authors. Capsule networks is the new and shiny neural network architecture that is challenging CNN, currently the king of the hill in computer vision.
It was an amazing and overwhelming experience for us to have been invited by our newfound friends Mike Richardson – who is a leading hardware design engineer just recently employed by True in Thailand and Julien Fredonie who is now head of Innovation at Honda.
Both these gentlemen worked very hard to make this a reality. Hundreds of people showed up from academia, It was an amazing collection from Academia, Start-ups and businesses and we were glad to have share our vision on not only what current Capsule Networks can provide but also what we are cooking in our labs.
The workshop was primarily about how Deep Learning industry is still advancing and what exactly do CNN (convolutional neural networks) do and cannot do. This very weakness of CNNs has been addressed by the authors of this new NN theory called Capsules Networks. We then also discussed some of our own findings on additional research around Manifold Learning and the use of PGM (Probabilistic Graph Models).
The lecture took about 6 hours to complete and it was very encouraging to notice that many learners stayed to the very end of the lecture. After the event we had several interesting conversations regarding the practical and real-world use of AI and Deep Learning in industry verticals.