An error occurred when I used maximization :
f1 = _ScalarObjective(name='f1', evaluator=f_1, maximize=True)

This is the error :
File "/home/ai/anaconda3/lib/python3.7/site-packages/desdeo_problem/Problem.py", line 718, in init
to_maximize = sum(to_maximize, []) # To flatten the list
TypeError: can only concatenate list (not "bool") to list

If you are not using the Jupiter notebook, and try to give the reference point within the after the first iteration you will get an error.

This is the code for getting the reference point :

while evolver.continue_evolution():

##getting the reference point:

for index in range(n_obj):
    pref_val = float(
        input("Please input the reference point: ")
        )
    refpoint[index] = pref_val
pref.response = pd.DataFrame([refpoint], columns=pref.content['dimensions_data'].columns)
plot= evolver.iterate(pref)
figure = animate_next_(
    plot[0].content['data'].values,
    figure,
    filename="river.html",
    generation=evolver._iteration_counter,
)

Here is the error that I get:
"Wrong preference object sent. Expected id = 1389300906.
Recieved id = 9506779031"

EMO to solve an multi-scenarioS-MOP example notebook does not work, because a file dependency notebook_extras/selected_scenarios_river(12.5.22).csv is missing from the repository.

When using surrogates, we need to access the actual objective functions to update them in online optimization problems.

More details: lets introduce a new term here as interactions. Each interaction is consists of a number of iterations, which is consists of a number of generations.

In each generation, we run an EMO algorithm.
In each iteration, we select some solutions to evaluate with original objectives and update the surrogates.
In each interaction, we select some solutions to evaluate them with original objectives and show them to the DM.

So since I really like the way the selection methods are handled, my suggestion is to do the same thing with the model management (if possible).

Other python versions are now supported.

Also, np.asarray should only have 1 input?

Colour the lines based on selected objective/variable/"distance" from the ideal point.

Check this out:
X.Blasco J.M.Herrero J.Sanchis M.Martínez: A new graphical visualization of n-dimensional Pareto front for decision-making in multiobjective optimization, Information Sciences, 178(20), 3908-3924, 2008, https://www.sciencedirect.com/science/article/pii/S0020025508002016

For example: Das-Dennis algorithm

Make sure that versions are mentioned only up to minor levels, not patch levels

figured it out!

Don't mind this.!

"ValueError: high is out of bounds for int32"

Explanation: https://stackoverflow.com/questions/36278590/numpy-array-dtype-is-coming-as-int32-by-default-in-a-windows-10-64-bit-machine

I think there should be a K-means clustering at the end of each iteration with S centers (S is the number of solutions that the DM wants to see) so that the DM does not have to see all the solutions.

"ideal point" repeated twice

EAs should treat positive values as good and negative values as bad to be consistent with the MCDM side.

This will enable support for more versions of python. But we need to replace the code for non-dominated sorting.

sometimes we only know the ideal point in advance and we have no idea about Nadir. However in the current implementation, one cannot use only one of them as input (check line 26-28 in Ppoulationl.py).

It should be split into two "if"s so we can just give one of the Ideal or Nadir points as input

Especially NSGA-III. Many components can be made much faster.

Some users may not be aware of what virtual environments are. Make the concept clearer. Provide some links perhaps, or code.