Should I always give output values ​​in the evaluate function?

[ubeydtalha]

First of all, I would like to thank everyone who contributed to this library. ^^

I'm developing a circuit optimization tool, simulating the input values ​​in the circuit and passing the output values ​​to pymoo.

But a numerical value does not always come from the circuit simulation, the circuit may give an error according to the individuals produced.

In this case, the value is "failed" and I cannot output this value.(out["F"])

In this case, is it sufficient to just terminate the function with exiting with "return"?

Nor can I define a constraint function and have it fail manually, as I have to use the MOEA/D algorithm.

[blankjul]

This looks like a hidden constraint to me with failed runs. The easiest fix is to wrap your function with try...except and set the objective value(s) to np.inf whenever it fails.
Let me know if this works for you.

[ubeydtalha]

I ended the function with return without giving any value to the purposes.but I don't know how true?
I will try what you said ^^

    def _evaluate(self, x, out, *args, **kwargs):

        meAbsPath = os.path.dirname(os.path.realpath(__file__))
        h = HSpicePy(file_name="amp",design_file_name="designparam",path=meAbsPath,timeout="")
        h.change_parameters_to_cir(LM1 = x[0], LM2 = x[1], LM3 = x[2],
                        WM1 = x[3], WM2 = x[4], WM3 = x[5],
                        Rb = x[6],Vb = x[7])
        asyncio.run(h.run_async())
        # h.run()
        f1_ = h.result["bw"]
        f2_ = h.result["gain"]
        
        if "failed" in [f1_,f2_]:
            return
            out["F"] = [0, 0]
            
        else:
            f1 = float(h.result["bw"])
            f2 = -float(h.result["gain"])

            out["F"] = [f1, f2]
```

[blankjul]

just set out["F"] = [np.inf, np.inf] when it fails.

[ubeydtalha]

Why can't we add out["G"] to the MOEA/D algorithm?

[blankjul]

the MOEA/D algorithm does not have any constraint handling (not part of the original paper)

[ubeydtalha]

Got it, thank you very much for your help.

[ubeydtalha]

just set out["F"] = [np.inf, np.inf] when it fails.

hello, when you do as you say, x values ​​start to be created in true-false, not float.

To prevent this, when I write return without giving out["F"] to the _eveluate function, I get the this error

[blankjul]

I would not no of any reason why setting "F", should impact the "X" of any iteration in the future. Have you figured it out?