[Comp-neuro] Postdoc/PhD Position in Recurrent Neural Network Architectures

Herbert Jaeger h.jaeger at jacobs-university.de
Mon Aug 31 11:06:46 CEST 2015

The department of Computer Science at Jacobs University Bremen 
(http://www.jacobs-university.de) announces a


(full-time, very competitive salary, 3 years, start December 1, 2015 or 

the MINDS research group of Herbert Jaeger 
(http://minds.jacobs-university.de). The position is created in the 
European Horizon 2020 integrated project "Neural Computing Architectures 
in Advanced Monolithic 3D-VLSI Nano-Technologies" (NeuRAM3).

The project aim of NeuRAM3 is to design and fabricate a large-scale, 
highly configurable neuromorphic microchip. Based on spiking, recurrent 
neural dynamics, the microchip is similar in some ways to IBM's recent 
TrueNorth neurochip 
(http://www.research.ibm.com/articles/brain-chip.shtml). However, our 
project aims significantly further than TrueNorth, and IBM is one of the 
consortium partners. In particular, novel 3-dimensional chip 
manufacturing technologies will enable a much more complex and much more 
flexibly configurable on-chip neural architecture.

Within this setting, the MINDS group develops *neural learning 
algorithms and architectures* which are tailored to the novel 
opportunities afforded by a 3-dimensional, spiking microchip. This line 
of research will be carried out in close collaboration with a consortium 
partner from computational neuroscience (Institute of Neuroinformatics, 
University Zurich and ETH Zurich).

Currently in machine learning approaches of deep learning are attracting 
much attention. In several ways the research in the MINDS group is 
*complementary to deep learning*. The latter is grounded in 
high-precision gradient descent methods for training and can devour 
extremely large training datasets to learn extremely complex tasks. The 
price paid is very long training times and a lack of online 
adaptability. In contrast, *we aim for very fast (even instantaneous) 
learning, online adaptability and extensibility, robustness against 
noise and computations with low numerical accuracy*. Such properties 
will be crucial for the algorithms running on next-generation chipsets 
for mobile devices – and they are also crucial for biological neural 
systems. The background philosophy of this research is framed by 
reservoir computing (http://minds.jacobs-university.de/ESNresearch) and 
conceptors (http://minds.jacobs-university.de/conceptors).

Eligible candidates must have a mindset inclined toward mathematical 
modeling, fluency in English, plus a solid background in at least one of 
the following areas:

•	dynamical systems theory,
•	signal processing and control,
•	machine learning,
•	computational neuroscience,
•	intelligent agent architectures.

Details about the position and the NeuRAM3 project will be gladly 
supplied by Herbert Jaeger (h.jaeger at jacobs-university.de) upon request. 
Please submit applications by email to Herbert Jaeger. Applicants 
passing an initial screening will be asked to supply two letters of 
reference. Applications will be processed as long as the job announmcent 
is posted online at http://minds.jacobs-university.de/jobs.

*Jacobs University Bremen* is Germany's only private university with a 
full range of academic programs from the natural and social sciences and 
engineering. Its student body is one of the most international 
conceivable: more than 100 nations are joined on our beautiful campus. 
Within its young history of only a little more than 10 years, Jacobs 
University has climbed to the top in many disciplines in Germany's most 
important university ranking 
(http://www.che-ranking.de/cms/?getObject=613&getLang=en). The language 
on campus is English.

Dr. Herbert Jaeger

Professor of Computational Science
Jacobs University Bremen gGmbH
Campus Ring
28759 Bremen, Germany

Phone (+49) 421 200 3215
Fax (+49) 421 200 49 3215
email  h.jaeger at jacobs-university.de


More information about the Comp-neuro mailing list