Lunch meeting on Artificial Intelligence

Together with HTSC and Center H&T, DSC/e hosted the first Artificial Intelligence lunch meeting on January 31. This meeting was intended for all interested TU/e researchers in this inspiring topic. In this highly visited event (all 50 seats were taken), Johan Lukkien (dean of the Mathematics & Computer Science department) gave some context and shared the current AI plans at the TU/e. He also presented the AI Landscape and invited all attendees to get involved in further shaping this initiative. The main menu consisted of three talks from different angels of approach on AI research:

AI fundamental research

Democratizing and Automating Machine Learning

Joaquin Vanschoren (DSC/e)

AI applied research

AI for Mobile Perception Systems

Gijs Dubbelman (HTSC)

AI human interaction research

AI and Human Interaction

Wijnand IJsselsteijn (H&T)

The meeting clearly served its goal to increase awareness about state of the art AI research and to build an internal network. Discussions after the presentations had to be kept short to give the podium to all presenters. A number of follow up meetings will be organized during 2018.


Democratizing and Automating Machine Learning
Dr. Joaquin Vanschoren (M&CS – Data Mining group)

Building machine learning systems remains something of a (black) art, requiring a lot of prior experience to select appropriate ML architectures (deep learning architectures or more traditional ML workflows) and their hyper parameters. To democratize machine learning, and make it easily accessible to those who need it, we need a more principled approach to experimentation to understand how to build machine learning architectures, and progressively automate this process as much as possible. First, we created OpenML, an open science platform allowing scientists to share datasets and train many machine learning models from many software tools in a frictionless yet principled way. It also organizes all results online, providing detailed insight into the performance of machine learning techniques, and allowing a more scientific, data-driven approach to building new machine learning systems. Second, we use this knowledge to create automatic machine learning (AutoML) techniques that learn from these experiments to help people build better models, faster, or automate the process entirely.

AI for Mobile Perception Systems
Gijs Dubbelman (EE - Video Coding & Architectures group)

In this presentation, an overview of AI-related projects of our High-Tech Systems Center is presented. Specific national and international inter-faculty projects on AI and robotics are high-lighted. 
The presentation then zooms in on the work of the Mobile Perception Systems research cluster of our EE department. Its research is a combination of Computer Vision, Deep Learning, and Sensor Fusion in the domain of intelligent and self-driving vehicles. Driving a vehicle is a very complex cognitive task and trying to mimic it with current AI techniques immediately brings forward their fundamental limitations. 

Therefore, it is an excellent use-case for both applied and fundamental research on AI. The presentation concludes with recent results and an outlook on future work.

Wijnand IJsselsteijn (IE&IS - Human Technology Interaction group)