Volume 464, Number 3, March IV 2007
|Page(s)||1167 - 1175|
|Published online||22 January 2007|
A new approach to the optimal target selection problem
South African Astronomical Observatory, Observatory, Cape Town, South Africa e-mail: [ed;bruce;heyden]@saao.ac.za
2 Mathematics Department, University of Cape Town, Rondebosch, South Africa
3 Department of Physics, University of Cape Town, Rondebosch, South Africa e-mail: firstname.lastname@example.org
4 iThemba Labs, Faure, Cape Town, South Africa
Accepted: 15 January 2007
Context.This paper addresses a common problem in astronomy and cosmology: to optimally select a subset of targets from a larger catalog. A specific example is the selection of targets from an imaging survey for multi-object spectrographic follow-up.
Aims.We present a new heuristic optimisation algorithm, HYBRID, for this purpose and undertake detailed studies of its performance.
Methods.HYBRID combines elements of the simulated annealing, MCMC and particle-swarm methods and is particularly successful in cases where the survey landscape has multiple curvature or clustering scales.
Results.HYBRID consistently outperforms the other methods, especially in high-dimensionality spaces with many extrema. This means many fewer simulations must be run to reach a given performance confidence level and implies very significant advantages in solving complex or computationally expensive optimisation problems.
Conclusions.HYBRID outperforms both MCMC and SA in all cases including optimisation of high dimensional continuous surfaces indicating that HYBRID is useful far beyond the specific problem of optimal target selection. Future work will apply HYBRID to target selection for the new 10 m Southern African Large Telescope in South Africa.
Key words: cosmology: large-scale structure of Universe / methods: statistical / surveys
© ESO, 2007
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.