-------------------------------
Chalmers University, Sweden
Contributed by: Jonas Sjöberg, jonas.sjoberg@chalmers.se
Ph.D Position on Nonlinear System Identification
Department of Signals and Systems (S2) at Chalmers University, Goteborg, Sweden
The project aims at developing methods for identifying nonlinear black box state space models where over-fitting is avoided. The approach to be studied is where an initial linear model is being used to initialize a nonlinear model by adding parameterized basis functions to the linear model. Starting with flexible models containing many parameters, methods for eliminating a substantial subset of them will be developed. A nice way to do this is to use L1-regularization in the criterion of fit but general algorithms to calculate the estimate, such as LASSO, only exist for models which are linear in the parameters. Hence, for state space models one needs to relax the original problem and this will be done by dividing the problem inrito to problems that are iteratively solved: 1) estimate the states and 2) using the states to estimate the nonlinear function describing the time update of the state vector and the function describing the output signal.
The working time of a Ph.D. student is mainly devoted to research. Undergraduate teaching duties, not exceeding 20% of the working time, may include supervision of MSc students.
Full-time temporary employment. The position is limited to a maximum of five years.
For further details and application submission, please go to
http://www.chalmers.se/en/about-chalmers/vacancies/?rmpage=job&rmjob=1363
Application deadline: 2013-06-20
Contributed by: Jonas Sjöberg, jonas.sjoberg@chalmers.se
Ph.D Position on Nonlinear System Identification
Department of Signals and Systems (S2) at Chalmers University, Goteborg, Sweden
The project aims at developing methods for identifying nonlinear black box state space models where over-fitting is avoided. The approach to be studied is where an initial linear model is being used to initialize a nonlinear model by adding parameterized basis functions to the linear model. Starting with flexible models containing many parameters, methods for eliminating a substantial subset of them will be developed. A nice way to do this is to use L1-regularization in the criterion of fit but general algorithms to calculate the estimate, such as LASSO, only exist for models which are linear in the parameters. Hence, for state space models one needs to relax the original problem and this will be done by dividing the problem inrito to problems that are iteratively solved: 1) estimate the states and 2) using the states to estimate the nonlinear function describing the time update of the state vector and the function describing the output signal.
The working time of a Ph.D. student is mainly devoted to research. Undergraduate teaching duties, not exceeding 20% of the working time, may include supervision of MSc students.
Full-time temporary employment. The position is limited to a maximum of five years.
For further details and application submission, please go to
http://www.chalmers.se/en/about-chalmers/vacancies/?rmpage=job&rmjob=1363
Application deadline: 2013-06-20
Zeeshan Rizvi
__._,_.___
| Reply via web post | Reply to sender | Reply to group | Start a New Topic | Messages in this topic (1) |
.
__,_._,___
No comments:
Post a Comment