About SPSANN Package - Githubissues

Laboratorio-de-Pedometria / spsann-package

Optimization of Spatial Samples via Simulated Annealing

8 stars 5 forks source link

About SPSANN Package #18

Closed yancong723 closed 4 years ago

yancong723 commented 4 years ago

In oder to finish my paper,I plan to use simulated annealing method and learn R about SPSANN. I have got some problems,however,it is very abstract when I read help document.

How users aggregate the objective functions into a single utility function？Such as several optimMSSD and optimMKV.
How understand chains and chain.length? How should I set the parameters in scheduleSPSANN.
What is the minmaxPareto function used for? What's the difference between minmaxPareto and aggregating the objective functions into a single utility function?

OrdinarySK commented 4 years ago

Can we communicate? Me too CUGer. Can we add a contact？

yancong723 commented 4 years ago

Of course !You can give me you cantact. We can communicate with wechat(18600792010),I am a Chinese.

OrdinarySK commented 4 years ago

Get

OrdinarySK commented 4 years ago

It is not find user to search your phone number in wechat. My Wechat (18571999690).

samuel-rosa commented 4 years ago

Hi @OrdinarySK and @yancong723. Sorry for the late reply. I have been dealing with health problems and will look into your question ASAP.

samuel-rosa commented 4 years ago

@OrdinarySK @yancong723 the only way for you to aggregate the two objective functions MSSD and MKV into a single utility function is by implementing the objective functions yourself and optimizing the sample configuration using optimUSER.

samuel-rosa commented 4 years ago

@OrdinarySK @yancong723 I agree that I need to improve the documentation. For you to fully understand the meaning of minmaxPareto please refer to the following papers:

Marler, R. T. & Arora, J. S. The weighted sum method for multi-objective optimization: new insights. Structural and Multidisciplinary Optimization, Springer Science + Business Media, 2009, 41, 853-862.
Marler, R. T. & Arora, J. S. Function-transformation methods for multi-objective optimization. Engineering Optimization, Informa UK Limited, 2005, 37, 551-570.
Marler, R. T. & Arora, J. S. Survey of multi-objective optimization methods for engineering. Structural and Multidisciplinary Optimization, Springer Science + Business Media, 2004, 26, 369-395.

samuel-rosa commented 4 years ago

@OrdinarySK @yancong723 again, I agree that I have to improve the documentation of this topic. For now, I recommend you to read the following papers to understand the meaning of chains and chain.length:

Ingber, L. Simulated annealing: practice versus theory. Mathematical and Computer Modelling, Elsevier BV, 1993, 18, 29-57.
Andrieu, C.; Freitas, N.; Doucet, A. & Jordan, M. I. An introduction to MCMC for machine learning. Machine Learning, 2003, 50, 5-43.

Please note that setting chains, i.e. the number of Markov chains has to be done by trial and error because it depends on your study setting. So, start with the default and see if the optimization stabilizes. Otherwise you'll have to increase the number of chains. However, with respect to chain.length, you can generally use the default value.