Learning Functions V1.4

Hi All,

Thanks for the useful software Uqlab. I have two questions

  1. Concerning the CMM learning function, is there a way in uqlab you can add more than one sample when used for reliability analysis. I noticed that when I used K-means clustering I could only add a sample at a time.

  2. I notice the uqlab supports EFF, U, CMM learning functions. Are there means to develop custom functions in uqlab or compare this learning functions with H,LIF,EI learning functions.

Thanks for all the support and hope to hear from you

Hi @aokoro,

  1. Yes you can add multiple points per iteration. For this you can use the .ALR.NumOfPoints option. Please have a look at the advanced ALR example for further information.

  2. As of now, this is not implemented in the module but it’s planned for next release. You could however have a look at the function uq_addEDPoints.m. It is relatively straightforward to add a new learning function. Say you would like to add LIF, then you can:

    1. Set ALROptions.ALR.LearningFunction = 'LIF' in the analysis options.
    2. Create an m-file uq_lf_lif.m in which you implement the learning function (you can copy uq_lf_u.m to get the right input/output structure )
    3. In uq_addEDPoints.m, add a case 'uq_lf_lif' in the first switch statement.
    4. And you’re good to go (In the next releasel, we plan adding the statement so that you don’t have to modify uq_addEDPoints.m).


PS: You can use which and edit commands in Matlab to find out the paths to the files I mentioned or to edit them (once the UQLab framework is started or if the UQLab folder is part of your Matlab paths already).

Thank you so much @moustapha for this detailed response. I will give this a try and let you know if I experience any difficulty.