Bio-Inspired Computation: Evolutionary Algorithm and Neural Networks

This project delves into the application of bio-inspired computation to solve the Traveling Salesman Problem (TSP), a classic combinatorial optimization challenge. The research is divided into two main parts. The first part focuses on implementing and comparing evolutionary algorithms, including variations in population size and mutation rates, to find the shortest distance between cities. The performance of the proposed algorithm is compared with other optimization algorithms, and the results are also analyzed using neural networks, particularly a Multilayer Perceptron (MLP) for classification. The second part involves a literature review of various bio-inspired algorithms, such as Ant Colony Optimization, bee colony, cuckoo search optimization algorithms and African Buffalo Optimization, and their application to the TSP. The research compares the strengths and weaknesses of these algorithms, highlighting their approaches to finding efficient solutions for the TSP. The conclusion summarizes the findings and emphasizes the potential of bio-inspired techniques in solving complex optimization problems.

Bio-Inspired Compution

Table of Contents
Introduction..............................................................................................................................1
Part 1A) – Evolutionary Algorithm........................................................................................1
Part 1B) – Neural Network Classification using MLP.........................................................3
Part 2 – Research on Articles about the Bio-inspired Algorithms......................................5
Conclusion.................................................................................................................................6
References.................................................................................................................................6

Introduction
The traveling salesman problem is an important combinatorial problem of optimization.
The goal of the problem is finding the shortest distance from one city which is the source to
another city which is the destination through n different cities. In this research, an evolution
algorithm is used for finding the shortest distance between n cities. This is used to solve tsp in
less amount of time. The issues that are included in implementing the evolution algorithm is
the encoding technique of the tour and the crossover algorithm that can be used for
combining the parent tours to create the child tour. The parameters that are used in this
algorithm are the population size and the number of generations. This research consists of
two parts. The first part consists of the implementation of the evolutionary algorithms and
comparing the performance of the proposed algorithm with the other optimization algorithms.
The sub division of the first part is the comparison of results using the neural networks. The
second part of the research consists of the integration part of the created Evolution algorithm
and the Multilayer Perceptron using neural networks.
Part 1A) – Evolutionary Algorithm
Evolutionary algorithm is used for solving the travelling salesman problem in the
fastest way. The evolutionary algorithm uses a kind of initial population. There are 6
parameters used in this research. They are population size, size of the group, Mutation,
Number of generations and the list of cities. The population size is initially taken as 5000.
The result for the 5000 population size is represented below.
Solution for Travelling Salesman problem

The population size is changed as 10000 and the mutation is changed as the 5%. Then
the result gets changed and the result is shown above.

After changing the size of the population as 100000 and the mutation as 10%, then the
result is represented above.
Part 1B) – Neural Network Classification using MLP
In this sub division of the part 1, the neural network is used for doing classification and
comparison between the other evolution algorithms.
Selection of Neural Network name and type

The above two windows represent the neural network classification. The first window
is used to specify the name and type of the neural network. The second window specifies the
input, hidden and output neurons. The input neuron is selected as 1, the hidden neurons is
specified as 5 neurons and the output neuron is specified as 1.
Output of the Neural Network Classification
The above diagram represents the neural network classification with 1 input and output
neurons and with 5 hidden neurons.
Comparison of the TSP solution results

The comparison results of the Traditional evolution algorithm and the proposed
evolution algorithm is tabulated above. The variations of the results of the various algorithms
are shown and tabulated above.
Part 2 – Research on Articles about the Bio-inspired Algorithms
The bio inspired algorithms such as Ant colony optimization, bee colony and cuckoo
search optimization algorithms are compared. The ant colony optimization algorithm (ACO)
is the technique that is used for solving the problems of computation which is used in
reducing the cost of the travelling from one city to another city. The proposed algorithm
mainly concentrates on the nearest neighbour node in the traveling salesman problem by
making the use of benchmarking (Pintea, Pop and Chira, 2017).
The paper proposed a new novel approach called Mosquito Host Seeking Algorithm
(MHSA). This is the algorithm which is inspired by the host-seeking behavior of the
mosquitoes to solve the traveling salesman problem. The MHSA algorithm is derived from
the new branch of the bio-inspired algorithms that can be used for solving the tsp problems.
This algorithm is used for describing the complex, high dimensional, high non-linear
behaviors of the entities. The simplicity of the MHSA algorithm is compiled with high
amount of flexibility. This approach is easily adaptable with large amount of the optimization
problems (SoleimanianGharehchopogh, Maleki and Farahmandian, 2012).
The article specifies a method called Swarm intelligence for solving the traveling
salesman problem. This method is considered to be the bio inspired approach which is used to
describe the behavior of the self organized approach to solve the travelling salesman problem.
This self organization methods include many number of mechanisms that ensures the

achievement of the global target and this is the fundamental of the social manner of the ant.
This approach is the result of four different components called randomness, multiplicity,
positive and the negative feedbacks of the system. The best case time complexity algorithm is
O(nlogn) and the worst case time complexity algorithm is O(n3) (V.V. and Y.A., 2017).
This paper specifies a new method of bio inspired algorithm called African Buffalo
Optimization. This algorithm is a new kind of meta heuristic algorithm which is derived from
the observation of the species of the African Buffalo. The results that are obtained from the
African buffalo optimization algorithm are very competitive in nature. These results are used
for solving the travelling salesman instances. Stagnationhandling by the use of updates of the
best buffalo for every iteration. The relative parameters can be used to make sure of the fast
convergence. This algorithm consists of only less than 100 lines of code in any type of
programming language (Odili and Mohmad Kahar, 2015).
This paper defines the comparisons of meta heuristic algorithms called two classical,
evolutionary algorithm called genetic and memetic and then 4 number of naturally inspired
algorithms called Ant colony, Bee colony and cuckoo search can be used to solve the
traveling salesman problem. This paper has observed the results of the various heuristic
algorithms. It says that the cuckoo search algorithm is best to solve the traveling salesman
problem (Gupta, 2013).
Conclusion
The traveling salesman problem is considered as the combinatorial problem of
optimization. The goal of the problem is to determine the shortest distance from one city
which to another city through n different cities. In this research, an evolution algorithm is
used for finding the shortest distance between n cities. This research consists of two parts.
The first part consists of the implementation of the evolutionary algorithms and comparing
the performance of the proposed algorithm with the other optimization algorithms. The sub
division of the first part is the comparison of results using the neural networks in the multi-
layer perceptron with input and output neurons. The second part of the research consists of
the report that specifies the research about the artciles on the bio-inspired algorithms.
References
Gupta, D. (2013). Solving TSP Using Various Meta-Heuristic Algorithms. International
Journal of Recent Contributions from Engineering, Science & IT (iJES), 1(2), p.22.

Odili, J. and Mohmad Kahar, M. (2015). Solving the Traveling Salesman’s Problem Using
the African Buffalo Optimization.
Pintea, C., Pop, P. and Chira, C. (2017). The generalized traveling salesman problem solved
with ant algorithms. Complex Adaptive Systems Modeling, 5(1).
SoleimanianGharehchopogh, F., Maleki, I. and Farahmandian, M. (2012). New Approach for
Solving Dynamic Traveling Salesman Problem with Hybrid Genetic Algorithms and Ant
Colony Optimization. International Journal of Computer Applications, 53(1), pp.39-44.
V.V., K. and Y.A., K. (2017). Bioinspired algorithm applied to solve the travelling salesman
problem.