Bio-inspired Evolutionary Algorithm Research Articles

Enhancing Cancer Classification through a Hybrid Bio-Inspired Evolutionary Algorithm for Biomarker Gene Selection

Enhancing Cancer Classification through a Hybrid Bio-Inspired Evolutionary Algorithm for Biomarker Gene Selection

Guiding the optimization of membraneless microfluidic fuel cells via explainable artificial intelligence: Comparative analyses of multiple machine learning models and investigation of key operating parameters

Membraneless microfluidic fuel cells (MMFCs) offer great potential for clean energy production, but their expense and tedious optimization process have limited their wider use. Machine learning (ML) algorithms have shown promise in improving the optimization of MMFCs, but the “black box” nature of these models has caused uncertainty and limited their adoption. To address this, we conducted the first study that implements the explainable artificial intelligence (XAI) approach to gain in-depth insights into multiple ML optimization models' predictions and the impacts of operating parameters on MMFCs' performance. Among 27 investigated models that are generated based on nine ML and three bio-inspired evolutionary algorithms, the combination of a decision tree and the particle swarm optimization algorithm is the most effective model that satisfied all criteria of high time efficiency (execution time = 3.194 s), accuracy (squared correlation coefficient = 0.999), and optimum power density identification (0.278 mWcm−2). The ML-optimized power density is 239.024% increased, which is 3.4 times higher than the average power density obtained without optimization. We further utilized the XAI model to comprehensively examine the impacts of operating parameters, providing valuable decision aids for the researchers, technicians, and engineers in manufacturing optimal power MMFCs at a significantly lower cost and faster rate.

SEARCH FOR OPTIMAL FUNCTIONS OF FUZZY SYSTEMS BASED ON BIOINSPIRED EVOLUTIONARY ALGORITHMS

Contemporary research in the field of creation and development of intelligent systems based on fuzzy logic is carried out mainly in the direction of developing highly efficient methods for their synthesis and structural-parametric optimization. In recent years, due to the intensive development of information technologies and computer hardware, bioinspired intelligent techniques of global search are quite promising for solving problems of synthesis and optimization of fuzzy systems, which include evolutionary and swarm methods, that simulate the processes of natural selection, as well as collective behavior of various groups of social animals, insects and microorganisms in nature. This paper is devoted to the development and study of a method of optimal membership functions search for fuzzy systems based on bioinspired evolutionary algorithms of global optimization. The obtained method allows finding the optimal membership functions of linguistic terms at solving the compromise problems of multicriteria structural optimization of various fuzzy systems in order to increase their efficiency, as well as to reduce the degree of complexity of further parametric optimization. In the proposed method for finding the global optimum of the problem being solved, the iterative procedures are carried out on the basis of combination of several different bioinspired evolutionary algorithms with subsequent analysis of the results obtained and the choice of the best variant of the membership function vector. The paper outlines the theoretical foundations and information model for the implementation of the computational step-by-step method for structural optimization of fuzzy systems, as well as presents various options for carrying out its search procedures. In particular, the features of the application and adaptation to the search problem to be solved of such bioinspired evolutionary algorithms as genetic, artificial immune systems and biogeographic are discussed.

Neuroevolution-based autonomous robot navigation: A comparative study

The field of neuroevolution has achieved much attention in recent years from both academia and industry. Numerous papers have reported its successful applications in different fields ranging from medical domain to autonomous systems. However, it is not clear which evolutionary optimization techniques lead to the best results. In this paper, multilayer perceptron (MLP) neural networks (NNs) are trained and optimized using four advanced bio-inspired evolutionary algorithms (EA). The algorithms are Multi-Verse Optimizer (MVO), Moth-flame optimization (MFO), Cuckoo Search (CS) and Particle Swarm Optimization (PSO). Each algorithm is equipped with two operators: evolutionary population dynamics and mutation, which impact on exploration and exploitation. Optimized MLPs are then used for the navigation of an autonomous robot. Accuracy and area under the curve metrics are used for the evaluation and comparison metrics. Moreover, two well-regarded gradient descent algorithms including Back propagation (BP) and Levenberg Marquardt (LM) are utilized to validate the results obtained by evolutionary-based MLP trainers. It is observed that MLPs developed using MFO are the most robust ones among MLPs trained using other evolutionary and gradient descent algorithms.

Diabetic Foot Risk Classification using Decision Tree and Bio-Inspired Evolutionary Algorithms

Diabetic foot complications are a burden to the Indian population which affects both financially and physically. The complications could be prevented if the risk of diabetic foot are detected well in advance before the peripheral nerves are damaged leading to amputation and limb loss. The quantification of severity plays an important role in timely intervention, delivery of appropriate treatment and prevention of amputation. This can be modeled as a classification problem where the risk category is stratified into different levels of severity. This paper is an approach to build such a system, capable of classifying the risk category of diabetic patients for suitable follow-up and care. Decision trees are used for the same with features selected using bio-inspired evolutionary algorithms like Particle Swarm Optimization (PSO), Genetic Algorithm (GA), Cuckoo Search (CS), FireFly (FF), Dragon Fly (DF) and Gravitational Search Algorithm (GSA). The overall accuracy is 77% but it identifies the low risk and high risk cases effectively with 97% and 89% respectively.

Parameter extraction of solar photovoltaic cells and modules using current–voltage characteristics

ABSTRACTThe significance of bio-inspired evolutionary algorithms has attracted many applications for obtaining best solutions to their optimisation problems in the past decades. This paper is about the application of one of these algorithms, namely, quantum particle swarm optimisation algorithm for parameter extraction of solar photovoltaic cells using current–voltage (I–V) characteristics. This algorithm has been used here to extract five parameters, namely, photocurrent, saturation current, series resistance, shunt resistance and ideality factor that influence the I–V relationship of single diode model solar photovoltaic cells. This approach has been validated for a cell and a module. Simulations using Matlab software have shown that the simulated I–V characteristics obtained using the extracted parameters have good agreement with the experimental I–V values. The reason for the interest taken in undertaking this work is to suggest a good and an accurate simulator for solar system designers.

Swarm Intelligence & System Identification: A Hybrid Discrete Jumping Frogs Algorithm For Optimal Input Design

The representativeness of the flight dynamics model has become a major concern in aviation industry since numerical simulation is now inevitable in aeronautics and widely used for every new aircraft program. Consequently, the requirements on its accuracy and reliability increase constantly. To build a flight dynamics model of acceptable quality, flight tests are designed and flown to adjust some relevant coefficients of the preflight aerodynamic model to the real A/C. But, the constraint of building more accurate models from shorter flight tests campaigns leads to revisit this identification process by designing optimal inputs since the aerodynamic parameter estimation is dependent on the quality of the inputs sent to the aircraft. Optimal Input Design (OID) for dynamical systems corresponds to complex and difficult optimization problems characterized by an infinite-dimensional search space combined with the presence of multiple local extrema. Even if signals can be parametrized using function basis, so that problem becomes of finite dimension, OID dedicated to flight dynamics identification still remains hard to solve. Indeed, flight tests protocol optimization must be tackled in its entirety (single/multiple input/s for one or several flight tests) which means that problem dimension can be high. Moreover, decision variables correspond pratically to real-valued discrete parameters due to the tabulated aerodynamic data structure of the model. This article describes some recent advances in the field of OID for flight dynamics identification through an original and innovative bio-inspired evolutionary algorithm based on the theoretical principles of the Jumping Frogs Optimization (JFO) technique.

Techniques for Modeling Simulation Environments for Modular Robotics

AbstractIn modular robotics, complex structures can be formed from basic modules to solve tasks which would be difficult for a single robot. The development of techniques for adaptation and evolution of multi-robot organisms is the subject of Symbrion project Levi and Kernbach (2010). In the project, the bio-inspired evolutionary algorithms are massively simulated prior to run them on a real hardware. It is crucial to evolve behaviors of the robots in a simulation, that is close to a real world. Hence, accurate and efficient representation of an environment in the simulation is needed. The robots learn simple motion primitives or complex movement patterns during many runs of the evolution. The learned skills are then used during experiments with a real hardware. In this paper, we present methods for building 3D model of a real arena using a laser rangefinder. The resulting 3D models consist of triangles. They can be constructed in various level of details using state-of-the-art methods for 3D reconstruction. We will show, how the size of the models influences the speed of the simulation.