In recent years, Convolutional Neural Networks (CNN) have been widely used for real-world applications in the field of computer vision. Their class-leading performance, however, depends heavily on the architecture used for a given problem. In most cases, the architectures are manually optimized by the researchers, a time-consuming process hard to achieve without prior knowledge of CNN. In this paper, we propose a new genetic algorithm for the optimization of the CNN architecture for a given image classification problem. This algorithm extends and refines existing research in the field, by allowing depth exploration, introducing a novel sequential crossover operator, using an incremental selective pressure schedule over evolution (favoring higher diversity in early generations) and by evaluating individual performances over the validation set with early stopping. The technique is validated in three image classification dataset, namely, CIFAR10, MNIST and Caltech256 datasets, which are widely used benchmarks for image classification algorithms. We evaluate the performance and total execution time over these datasets, and compare our results with those achieved by the best genetic methods published so far. In all cases, we achieve better results in terms of test accuracy, consistently over different datasets, while remaining in the same orders of magnitude of execution time of the fastest approaches.

In the field of engineering, complex problems often arise that require solutions. The implementation of specific algorithms to tackle these problems becomes an essential part of this area, and optimizing these algorithms plays a crucial role in achieving good results with the available resources. The Vehicle Routing Problem (VRP) has been a central topic in distribution and logistics for decades. New VRP models and tools are continuously developed to address the challenges of modern logistics. The Energy Minimizing Vehicle Routing Problem (EMVRP) is a VRP model where the objective is to minimize the total amount of energy consumed by a fleet of vehicles. The VRP literature has focused on solving the problem using a variety of approaches and techniques, including exact methods, heuristics, metaheuristics, and hybrid algorithms. Hybrid algorithms combine different techniques, such as heuristics and metaheuristics, to obtain more effective solutions. This work presents the implementation of four hybrid algorithms to tackle the EMVRP. These algorithms combine machine learning clustering techniques, such as K-Means and K-Medoids, with metaheuristic approaches inspired by ant colony systems. The proposed algorithms are: Free Ant + K-Means, Free Ant + K-Medoids, Restricted Ant + K-Means, and Restricted Ant + K-Medoids. The Free Ant and Restricted Ant algorithms were subjected to testing using different instances from CVRPLIB, and the obtained results are promising compared to other metaheuristic techniques described in the literature. However, these results also indicate a wide scope for experimentation and algorithm tuning.

The Set Covering Problem consists in finding a subset of columns in a zero-one matrix such that they cover all the rows of the matrix at a minimum cost. To solve the Set Covering Problem we use a metaheuristic called Binary Cat Swarm Optimization. This metaheuristic is a recent swarm metaheuristic technique based on the cat behavior. Domestic cats show the ability to hunt and are curious about moving objects. Based on this, the cats have two modes of behavior: seeking mode and tracing mode. We are the first ones to use this metaheuristic to solve this problem; our algorithm solves a set of 65 Set Covering Problem instances from OR-Library.

The set cover problem, belongs to the branch of combinatorial optimization problems, whose complexity is exponential theoretically established as NP-complex problems. Consists in finding a subset of columns in a matrix of zeros and ones such that cover all rows of the matrix at a minimal cost. In this work, the problem is solved by binary Firefly algorithm, based on the flashing behavior of fireflies, using binary representation. A firefly produces a change in brightness based position between the fireflies. The new position is determined by the change in the value of the old position of the firefly, but the number of the new position is a real number, we can solve this problem with the function tanh binarization compared with a random number generated uniformly distributed between 0 and 1. The proposed algorithm has been tested on 65 benchmark instances. The results show that it is capable of producing solutions competitivas. virtualización; ultrasecuenciación genetics.