"The Deep Learning Algorithms used in Medical Diagnostics and Self-Driving Cars"
Senior Deep Learning Research Engineer
on Monday 19 June, 10.00-11.00am. (Talk) 11.00-12.00am. (Lab, to learn technical stuff)
at room 106, CSIM Building.
For the first part of presentation, I will focus on the DREAM Challenge breast cancer recognition contest that I've been working on. I will share technical informations about the proposed architecture of a winner (Yaroslav Nikulin).
In the second part, I will talk about the current state of self-driving cars and overview of latest deep learning algorithms used by them, adding how they are optimized to achieve the best performance. And he will also show a practical example of building object detection algorithm in Keras, together with network activations visualization.
Bio: Marek started to intensively learn Machine Learning in 2012, participating in multiple competitions held by NASA, Harvard and DARPA with excellent results. His unique skills improved the International Space Station rotation algorithm to collect enough power from solar panels while preventing their overheating. Has been honored by Harvard professors and the Director of Advanced Exploration Systems at NASA, who signed the official Letter of Recommendation. From the first summer during university course, Marek taken internships at NVIDIA and Microsoft. Post graduation, he has been invited to work on a critical project for NVIDIA - TensorRT - in Switzerland and quickly become recognized as a senior R&D engineer. At TensorRT, he is responsible for developing multiple features, including leading the development of the fast batch=1 inference algorithm, that reduced the deep learning inference time by half for self driving cars.
1. Talk [40-60mins]
Deep learning for self-driving cars and medical diagnostics.
2. Lab (Optional, for people who want to learn technical stuff) [40-60mins]
Introduction to Neural Networks with Keras
- Building Neural Networks
- Building an Object Detection Algorithm
- Visualizing Network Activations
3. Brief Networking