0

I use the sparse Gaussian process for regression from Rasmussen. [http://www.tsc.uc3m.es/~miguel/downloads.php][1]

The syntax for predicting the mean is:

[~, mu_1, ~, ~, loghyper] = ssgpr_ui(Xtrain, Ytrain, Xtest, Ytest, m);

My question is, the author states that the initial hyper parameter search condition is different for different iterations, hence the results of the model is different from every iteration. Is there any way to ensure that the best initialization or seed condition is set to have good quality hyper parameters for best predictions and reproducible results?

Arvind Kumar
  • 66
  • 8
  • As this question is mostly about the method and not programming, have you considered migrating the question over to stats.stackexchange.com? – mikkola Nov 28 '15 at 09:01
  • @mikkola : it's relevant to programming of course. ( setting the Seed) – Arvind Kumar Nov 28 '15 at 09:38
  • 1
    Right, sorry for the confusion. Trial and error comes to mind - but more importantly, doesn't fine tuning the model for a certain data set in a sense defeat the purpose of regression modelling? You will find a very nice model for your training and validation data, but I think there's a decent chance you will overtrain and lose generalisation ability. – mikkola Nov 28 '15 at 10:12

1 Answers1

1

In order to obtain the same predictions every time, it is possible to set the seed by stream = RandStream('mt19937ar','Seed',123456); RandStream.setGlobalStream(stream);

However, there is no standard procedure to set the best seed. Doing so will lead to over fitting of the model as we are giving too much of ideal conditions to fit the training data as quoted by @mikkola

Arvind Kumar
  • 66
  • 8