Add multiple multi-objective optimizers for solution composition (#200)

[heidmic]

• Add MOSolutionFitness and implement non_dominated_sorting

• Fix MOSolutionFitness and Implement NSGAII

• Add stochastic Sampling and test example

• Implement SPEA2

• Add MOOSolutionComposition ABC.

• Add method to get pareto front.

• Add Pareto front logging

• Add Hypervolume metric

• Add Spread metric

• Add reference based selection

Added Das-Dennis reference point generation, an association function and the niched selection operator from U-NSGA-III.

• Add U-NSGA-III

• Add dynamic worst point estimation

The worst points is now not taken from the fitness function but instead from the worst observed objective values in each dimension.

• Add DiversitySolutionSampler

A diverstiy based solution sampler that calculates weights based on crowding distances

• Add BetaSolutionSampler

A BetaSolutionSampler which represent a beta distribution over the Pareto Front.

Points can optionally be projected onto the normal vector simplex, and onto the x-axis (currently works in only 2 dimensional objective functions)

• Add Warm Start to TSSC

• Add soo plotting to example_5

• Implement Hypervolume based early stopping in MOSC

*Implement early stopping
*Retain best pareto front
*Fixed critical bug in spea2 that always
caused termination after one iteration.

• Sc early stopping (Add commit message guide to README #10)

• Implement Hypervolume based early stopping in MOSC

*Implement early stopping
*Retain best pareto front
*Fixed critical bug in spea2 that always
caused termination after one iteration.

• Add SimpleMOSolutionFitness

• Add MOO SC test

• Adapt to SupRB MOO SC

---

[github-actions]

⚠️ Pylint job contains errors or warnings ⚠️

Please check the report here.
(The PR will still pass the pipeline)

[github-actions]

⚠️ Pylint job contains errors or warnings ⚠️

Please check the report here.
(The PR will still pass the pipeline)