Wednesday, December 7, 2016

[DMANET] PhD Candidate ‘Parameterized complexity of approximate Bayesian inferences’

PhD Candidate 'Parameterized complexity of approximate Bayesian inferences' (1.0 FTE)

Faculty of Social Sciences, U. Nijmegen, the Netherlands
Vacancy number: 24.49.16
Application deadline: 1 February 2017

Responsibilities

The recent Predictive Processing account in neuroscience postulates that the brain continuously predicts its future inputs, using generative models relating hypothesized causes to perceived effects. It entails that the brain actually implements some form of approximate Bayesian inference, combining prior expectations with newly arriving information to make sense of its inputs. Bayesian computations, however, are computationally intractable for situations of real-world complexity, even when approximated. This is in marked contrast to the efficiency of inference and learning as done by the brain in practice. The objective of this project is to resolve this paradox by studying the fundamental properties of approximate Bayesian inferences. The vital question we seek to answer is 'under what biologically plausible situational constraints can approximate Bayesian inference be rendered tractable?' The approach we use to address this question is the mathematical theory of parameterized complexity analysis, algorithm design, and computer simulation.

You will explore the boundaries between tractable and intractable (parameterized) sampling algorithms, contribute to the mathematical tools to assess such algorithms, and analyse the computations postulated in the Predictive Processing account. You will also contribute to our Artificial Intelligence educational programme, in particular by supervision of student projects and assistance in lab courses. In addition, you will be eligible to follow courses at the IPA and SIKS research schools. You will be part of the Donders Graduate School for Cognitive Neuroscience.

Work environment

The Donders Institute for Brain, Cognition and Behaviour focuses on state-of-the-art cognitive neuroscience, using a multidisciplinary approach, and offers excellent lab and neuroimaging facilities, PhD supervision and courses, and technical support.

The project is embedded in the Donders research theme "Perception, Action and Control" and will be supervised by Dr Johan Kwisthout. You will join the Computational Cognitive Science research group, headed by Dr Iris van Rooij, at the Donders Centre for Cognition (DCC). The DCC is part of the Faculty of Social Sciences, one of the largest faculties at Radboud University in Nijmegen. The institute offers excellent research facilities and study programmes that rank among the best in the Netherlands. The candidate will also maintain regular research contacts with the Algorithms and Complexity group, headed by Prof. Hans Bodlaender, at the Department of Information and Computing Sciences at Utrecht University.

The Donders Institute is an equal opportunity employer, committed to building a culturally diverse intellectual community, and as such encourages applications from women and minorities. The Radboud University offers a parental leave scheme and day care on campus.

What we expect from you

an MSc degree in computer science, mathematics, artificial intelligence, cognitive science, cognitive neuroscience, or a related field;
strong motivation to conduct a conceptually challenging theoretical research project in a highly interdisciplinary context;
you have strong formal and analytical skills and are fascinated by the foundations of computation as well as how the brain actually realizes computations;
knowledge of parameterized complexity theory and/or Bayesian networks would be a strong bonus.

What we have to offer

employment: 1.0 FTE;
in addition to the salary: an 8% holiday allowance and an 8.3% end-of-year bonus;
the gross starting salary amounts to €2,191 per month, and will increase to €2,801 in the fourth year;
you will be appointed for an initial period of 18 months, after which your performance will be evaluated. If the evaluation is positive, the contract will be extended by 2.5 years;
this PhD position at the Donders Centre for Cognition is a 0.9 FTE research and 0.1 FTE teaching appointment;
you will be classified as a PhD Candidate (promovendus) in the Dutch university job-ranking system (UFO).

Other Information

The intended start date of the project is April 2017.

For more information about this vacancy, please contact:
Dr. Johan Kwisthout, project leader
Telephone: +31 24 3655977
E-mail: j.kwisthout@donders.ru.nl
Are you interested?

You should upload your application (attn. of Dr. Johan Kwisthout) exclusively our vacancy website: http://www.ru.nl/werken/details/details_vacature_0/?recid=593059# Your application should include (and be limited to) the following attachment(s):

Cover letter
CV, including two references
list of grades

Please make sure that all documents are in English.

**********************************************************
*
* Contributions to be spread via DMANET are submitted to
*
* DMANET@zpr.uni-koeln.de
*
* Replies to a message carried on DMANET should NOT be
* addressed to DMANET but to the original sender. The
* original sender, however, is invited to prepare an
* update of the replies received and to communicate it
* via DMANET.
*
* DISCRETE MATHEMATICS AND ALGORITHMS NETWORK (DMANET)
* http://www.zaik.uni-koeln.de/AFS/publications/dmanet/
*
**********************************************************