PhD student - Developing theory of Reservoir computing for sensing applic.

Information about the project
Reservoir computing is a novel paradigm of computation. A reservoir computer consists of a dynamical system with sufficiently complex dynamics, and a read-out layer. An arbitrary input can drive the system to a specific region of the configuration space. In this way the system performs a transform of the input. The computation consists of  “inspecting” which part of the configuration space has been visited under a given input. This inspection is done by the read-out layer. In principle, such systems can process information without prior training, if they fulfil a known set of rather generic conditions.

The project aims to develop thorough theoretical understanding of possible use of reservoir computing for advanced sensing applications.  The concept is entirely novel, applicable to a broad range of sensing situations, and could be implemented for many sensor types.  If proven successful, this will be an embryo for the development of a new sensing technology, and for novel uses of the existing ones. As an illustration, one can envision applications  in various fields of bio-sensing, bioengineering, or medicine, e.g. an early disease detection by monitoring selected ion concentrations.

Major responsibilities
The work will be done in a close collaboration with several experimental groups that are working on an experimental prototype in the context of ion concentration characterization, and will consist of two major parts (a) the fundamental theoretical research, and (b) applied theoretical research.

(a) The fundamental theoretical development aiming at in depth theoretical investigation of the feasibility of the reservoir based sensing concept: (a1) Develop generic sensing model of computation. (a2) Generalize the model to the ion-concentration sensing context. (a3) Develop a theoretical model of the sensing-reservoir dynamics in the context of ion-concentration sensing with the following components: memristor networks, organic transistors, lipid flow memristors, photoactive elements, or silicon nanowires.

(b) Dissemination of the theoretical results to, and the use of the experimental results from the experimental groups: (b1) Develop necessary simulation tools in close synergy with experimental groups: software development, experimental data analysis, model parameterization using the experimental data. (b2) Provide guidelines for the experimental activities regarding the device design (to improve sensing capacity). (b3) Suggest suitable sensing scenarios.

Position summary
Full-time temporary employment. The position is limited to a maximum of four years.

Qualifications
To qualify as a PhD student, you must have a master's level degree corresponding to at least 240 higher education credits in the field of theoretical physics or other strongly related areas. A suitable candidate will likely come from disciplines such as complex dynamical systems or non-equilibrium statistical physics. An experience with modeling dynamical systems is required, and also experience with algorithmic development and implementation of algorithms (software development). In general, the successful candidate should have good problem solving skills, and good analytic and programming skills.

In addition, the following qualifications are (1) compulsory, (2) a strong bonus, or (3) will likely boost the candidate’s productivity on the project:

(1)  A candidate should have a good operational knowledge and understanding of the following: equations of motion, observables, stable vs unstable states, ergodic versus non-ergodic systems.

(2) An experience with system level modelling or multi-scale modeling, or experience with Markov processes: the master equation, stochastic simulation techniques (e.g. the minimal process method or other variants).

(3) An interest, and possibly some experience, in computer science (e.g. familiarity with some basic ideas from Theoretical Computer Science; the model of computation, the expressive power of the model, etc).

Application deadline: September 30, 2015

Important note: Interview invitations will be sent by October 9, and interviews will be conducted very soon after that, October 12-16. Please make sure to plan accordingly.


Background information: Some information about the theoretical approaches that will be used in the project can be found in
1. Lukoševicius, M. and H. Jaeger, Reservoir computing approaches to recurrent neural network training. Computer Science Review, 2009. 3(3): p. 127-149.
2. Kotomin, E. and V. Kuzovkov, Modern aspects of diffusion-controlled reactions : cooperative phenomena in bimolecular processes. Comprehensive chemical kinetics, ed. C.H. Bamford and C.F.H. Tipper. Vol. 34. 1996 Amsterdam Elsevier.

The first reference provides a gentle introduction to reservoir computing, which will be used to handle information processing aspects of the project.  The second reference provides an example of the basic theoretical modelling that might be done in the project. Both strongly feature the concept of the dynamical system.

For questions, please contact:
Zoran Konkoli, MC2, 
zorank@chalmers.se, 
+46 31 7725480