Shima Shafiee

Differential evolution, evolutionary strategy, and evolutionary programming are all viable possibilities. The technique should give an efficient solution for bigger problems than standard integer programming approaches for an acceptable GA design.

The methodologies were evaluated, and numerous experimental findings suggest that memetic algorithms outperform genetic algorithms for such NP-Hard combinatorial problems.

Python is one of the most popular platforms for genetic programming, with a plethora of interesting modules for genetic algorithms and good charting capabilities. Pyvolution, deap, pySTEP, PyRobot, DRP, and other libraries are among the most popular.

If you are looking for evolutionary algorithms, you can take a look at

1. Particle Swarm Optimization (PSO)

2. Ant Colony Optimization (ACO)

3. Differential evolution (DE)

Hello dear Qamar Ul Islam

Your answer is complete and useful

I will definitely check the proposed packages

Thanks a lot

DEAR Sikder Tahsin Al-Amin

Thanks a million for sharing your proposed algorithms

GOOD LUCK

Dimensionality reduction techniques are commonly used for feature construction. Traditional methods like PCA have been widely used for this, but nowadays autoencoders are increasingly popular. Autoencoders are a type of neural network, and can construct complex features either from existing feature sets or, more commonly, from raw data. The disadvantage of autoencoders in relation to genetic programming is that the resulting features are less interpretable.

Note that differential evolution, PSO etc (mentioned in previous answers) are alternatives to genetic algorithms; they're not alternatives to genetic programming, which is a different thing. Genetic programming evolves programs (typically represented as trees, but not always), whereas a genetic algorithm is a general-purpose optimiser. Genetic programming does actually use a genetic algorithm, but people have also created versions based around other optimisers, including PSO.

Hello Michael Lones

Thanks a lot for your explanation

Hi Shima,

Instead of wondering about what alternative algorithms you could use for any single problem, you might be inclined to adopt the more general approach of searching for the right combination of lower-level heuristics to best tackle your problem first. By that, I mean of course, hyper-heuristic methodologies.

Given the well-established limitations that generic metaheuristic algorithms have by dint of the NFL theorem, a hyper-heuristic approach is your best option moving forward.

Good Luck