Eindhoven University of Technology (TU/e) has a vacancy for a
PhD position on Deep Learning for High Tech Systems and Materials
(Project P4 within the Dutch Efficient Deep Learning program) within the Electronic Systems group, department of Electrical Engineering.
Deep Learning in contextDeep learning has dramatically improved the state-of-the-art in object detection, speech recognition, robotics, and many other domains. Whether it is superhuman performance in object recognition or beating human players in Go, the astonishing success of deep learning (DL) is achieved by deep neural networks trained with huge amounts of training examples and massive computing resources. Although already applied successfully in academic use cases and several consumer products (e.g. automatic language translation), its data and computing requirements pose all kind of efficient DL challenges for further market penetration like less time for training, low energy consumption, and the use on small embedded processing platforms as used in many systems.
The Efficient Deep Learning (EDL) programThe EDL program combines the fields of machine learning and computing: both disciplines are already strong in the Netherlands and now connected by 7 Dutch academic institutes and more than 35 other (industrial) partners in- and outside the Netherlands. The EDL program contains 7 use case driven EDL research projects: P1) DL as a service, P2) Reconstruction, matching and recognition, P3) Video analyses and surveillance, P4) High tech systems and materials, P5) Human and animal health, P6) Mobile robotics, and P7) DL platforms. Common goal for all 7 EDL projects is to significantly improve the applicability of DL, among others by creating data efficient training, and tremendously improving computational efficiency, both for training and inference.
The PhD positionPartners in EDL-P4 are the Dutch universities TU/e, UvA and VU, and also Thermo Fischer, Qualcomm (Scyfer), NLeSc, ASTRON and SURFsara. The PhD candidate will work, in the Electronic Systems group
www.es.ele.tue.nl/ml-at-es.php at the TU/e, on the development of energy and resource efficient DL algorithms and circuits for high performance embedded systems. This involves system modeling, and development of advanced training algorithms that work well with limited amounts of labeled data and are able to generate confidence bounds for the predictions made. He/she will work on implementations to accelerate DL by programmable hardware (e.g. FPGAs, GPUs).
The specific DL use case concerns
the application of DL for high-performance electron microscopy, in close cooperation with Thermo Fischer,
https://www.fei.com. Electron microscopy enables e.g. the study of biological tissue properties and cell structures, to obtain in-depth knowledge on disease properties and possible treatments.
He/she will also cooperate with the Amsterdam Machine Learning Lab (AMLAB) of the University of Amsterdam,
amlab.science.uva.nl, especially for Methods for semi-supervised learning and active labeling, where the goals are to use both labeled and unlabeled data in training a classifier, to train classifiers when the number of (labeled) examples is small and to detect bad labels and to suggest informative examples to be labeled.</p