Add inequality constraint of two different genes

[terrancelu92]

Hello, may I know how to add inequality constraint of two different genes. For example, I have a gene space gene_space=[{'low': 0, 'high': 16}, {'low': 0, 'high': 24}]. I would like to add the constraint of gene1 < gene2. May I know if there is a way to do it in PyGAD. Thank you!

[ahmedfgad]

@terrancelu92, this is definitely a good question and I am also thinking of adding constraints across the genes.

Definitely this can be supported in PyGAD. It is about controlling the values of these genes when:

• Creating the initial population.
• Applying crossover.
• Applying mutation.

Even if it is easy to add constraints like gene1 < or > gene2, I am looking to supporting a broad way to add conditions between the different genes. For example: $$gene_1*4+2 <= 2^{\dfrac{\sqrt{gene_2}}{5}}$$

I hope we can find a solution soon.

[terrancelu92]

@ahmedfgad ,Thank you for the reply! Could you please advise on where I need to revise based on the latest source code to implement the simple constraint e.g. gene1 < gene2?

I am now applying a large penalty on the objective function if the constraint is violated. However, by doing so I fear that the optimizer will focus more on avoiding the large penalty instead of optimizing the rest of the objective function.

[ahmedfgad]

Assuming you are going to use random mutation, then you need to edit this function:

GeneticAlgorithmPython/pygad.py

Line 2060 in 7f2b1f4

def mutation_randomly(self, offspring):

While generating the new value for the gene, make sure that its value meets your condition.

[123fgh123]

Assuming you are going to use random mutation, then you need to edit this function:

GeneticAlgorithmPython/pygad.py

Line 2060 in 7f2b1f4

def mutation_randomly(self, offspring):
While generating the new value for the gene, make sure that its value meets your condition.

Dear ahmedfgad,

Could you please give an example on how to edit function "mutation_randomly" to add a constraint e.g. gene1 < gene2?

Many thanks for your time.

Best regards