Congjian Wang - INL comments

Results 138 comments of


                                            Congjian Wang - INL

Mattdon/new melcor interface

@mandd Could you create an issue for this PR? In addition, could you update the description for this PR? I will review it this week.

[DEFECT] Optimizers have no way to handle failures

@aalfonsi FYI, we would like to resolve this issue, but it is not on our top priority list right now.

[DEFECT] Optimizers have no way to handle failures

@PaulTalbot-INL @mandd @dylanjm @Jimmy-INL Let's plan have a design meeting for this issue.

[DEFECT] Undefined or incorrectly defined constrain functions are not checked until second iteration of optimization.

@mandd @Jimmy-INL @JunyungKim This is another related issue, please make a plan and address it in the near future.

Model metadata to include output directory information

@aalfonsi I'm sorry there are other priorities that we need to work on. I would say I can start to look at this problem next week.

[DEFECT] Metropolis Sampler likelihood must be an output of the model

@aalfonsi: It seems to me both can be achieved using external model structure using the explicit expressions. If you want to using MCMC to generate samples and then use them...

[DEFECT] Metropolis Sampler likelihood must be an output of the model

@aalfonsi I think to support Function is useful. Regarding Distribution, in your example, the likelihood will be a predefined distribution, it is not clear to me what is the inputs...

[DEFECT] Metropolis Sampler likelihood must be an output of the model

@aalfonsi I think this is more a feature request.

[DEFECT] Gradient Descent - SPSA does not check for explicit constraints

@PaulTalbot-INL @dylanjm Please consider to resolve this issue. Thanks for your help.

[DEFECT] A lot of tests that uses constraint return Floats as return arg of function.

@JunyungKim Could you update the manual to resolve the confusion? and add a check in the GA code to make sure the constrain returns are float instead of boolean?