Machine learning enables computers to mimic human behaviour by teaching them historical data and knowledge about possible future events. This section will examine fascinating machine learning methods such as Crossover, edge recombination operator, and fly algorithm.

Crossover

Crossover, also known as recombination, is a genetic operator used in genetic algorithms and evolutionary computation to merge the genetic information of two parents to produce new offspring. It is comparable to the crossover during sexual reproduction in biology and represents one approach to stochastically making new solutions from an existing population. Cloning a current solution, which is comparable to asexual reproduction, is another way to create new solutions. Usually, new solutions are mutated before being introduced to the population. Moreover, each genetic representation can be paired with various crossover operators, and various evolutionary computation methods may employ different data structures to store genetic data. For example, bit arrays, real number vectors, and trees are common data structures that we can merge again via crossover.

Only some conceivable chromosomes represent a viable solution in some genetic algorithms. It is sometimes possible to utilise specific crossover and mutation operators that are made to follow the problem's requirements. An ordered list of cities, for instance, could be used by a genetic algorithm to indicate a solution path while tackling the travelling salesman problem. Only if the list includes every city the salesman needs to visit does such a chromosome offer a workable solution. The above crossings frequently produce chromosomes that defy that restriction. Therefore, different crossover operators are necessary for genetic algorithms that aim to optimise the ordering of a given list to prevent the generation of false positives.

Edge recombination operator

An adjacency matrix, which lists each node's neighbours in any parent, is the foundation for the edge recombination operator. For instance, in a travelling salesman problem like the one shown, the node map for the parents CABDEF and ABCEFD is generated by starting with the first parent, let's say 'ABCEFD,' and noting all of its close neighbours, including those that wrap around the end of the string.

It is a strategy for crossing over permutation (ordered) chromosomes. It aims to introduce the fewest pathways possible. In situations like the travelling salesman, adding a stray edge between two nodes is usually very detrimental to a chromosome's fitness. The idea is to exploit as many existing edges, or node connections, as possible to generate children. Edge recombination frequently outperforms PMX and Ordered Crossover, but it is typically more time-consuming to compute.

Fly algorithm

The Fly Algorithm is a cooperative coevolution strategy based on the Parisian method. In 1999, the researchers created the Fly Algorithm to use evolutionary algorithms for computer stereo vision. In contrast to the traditional image-based approach to stereovision, which extracts image primitives and then matches them to obtain 3-D information, the Fly Algorithm is founded on directly exploring the scene's 3-D space. A fly is a three-dimensional point specified by its coordinates (x, y, z). Once a random population of flies has been created in a search space corresponding to the cameras' field of view. Here, the fitness function employs the grey levels, colours, and textures of the fly's projections.

Reconstruction for emission Tomography in nuclear medicine is another application field of the Fly Algorithm. The Fly Algorithm has been effectively implemented in single-photon emission computed tomography and positron emission tomography. Each fly is considered a photon emitter, and the similarity between the simulated and actual sensor illumination patterns determines its fitness. Within this application, the fitness function has been redefined to incorporate the novel idea of "marginal evaluation." Here, an individual's fitness is determined by their (positive or negative) contribution to the quality of the global population. It is based on the cross-validation leave-one-out concept. A global fitness function analyses the population's quality as a whole; the fitness of an individual (a fly) is only then determined as the difference between the global fitness values of the population with and without the specific fly whose individual fitness function must be evaluated. Each fly's fitness is regarded as its "degree of confidence." During the process, it is employed to modify the individual footprint of the fly, utilising implicit modelling (such as metaballs). As a result, it creates smoother and more precise outcomes. Furthermore, recently, it has been employed in digital art to make visuals resembling mosaics or spray paint.