Soft Computing Techniques and itsApplications

International Journal of Scientific Research and Review Volume 07, Issue 05, May 2019 ISSN No.: 2279-543X UGC Journal No.: 64650 Soft Computing Techniques and itsApplications Kamal Saluja, Prof. (Dr.) Jatinder Singh Bal Research Scholar, I. K. Gujral Punjab Technical University, Jalandhar, Punjab, India, Professor, Universal Institute of Engineering & Technology, Lalru, Punjab, India Abstract Soft computing, unlike traditional computing, fits prediction models and provides answers to real complex problems. Unlike hard computing, soft computing tolerates great, uncertainty, partial truth and prediction. In fact, the model for soft computing is the human mind. Technologies such as fuzzy logic, genetic algorithms, artificial neural networks, machine learning and expert systems are part of soft computing. The theory of soft computing, introduced in the 1980s, is now a major area of research in all areas. Because of its low cost and high performance, soft computing is now successfully used in many home, commercial and industrial applications. The field of application of soft computing is increasing today. This article outlines the current state of Soft computing technology and explains the latter advantages and disadvantages to traditional computing technology.. Keywords: Soft computing; neural network, genetic algorithms; fuzzy logic; ANFIS. 1. Introduction One of the problems with conventional control systems is the inability to accurately describe complex equipment with mathematical models. Therefore, it is difficult to control them in such a way. Soft computing, on the other hand, is concerned with estimating partial problems, uncertainty, and complex problems to solve. A pioneer in fuzzy logic [1], the guiding principle of "soft computing" is to exploit credibility, uncertainty, and tolerance to partial truth. "Soft computing has become popular for features such as intelligent control, non-linear programming, support, decision support, and collaborative research has attracted interest for people with different backgrounds. It is difficult to use traditional control system technology to control the increasing complexity of modern machines. For example, many non-linear, time-varying plants can not be easily controlled and can be frozen for long periods using conventional techniques. One reason for this difficulty is the lack of an accurate model that describes the plant. Soft computing has proven to be an effective way to control such complex installations. Zadeh stated that soft computing is not a single method but a combination of several methods, such as fuzzy logic, neural networks, and genetic algorithms. All these methods are not competitive, but they fit together and can be used together to solve a given problem. Soft computing is the use of confusion and uncertainty in the decision making process to solve complex problems. 756 International Journal of Scientific Research and Review Volume 07, Issue 05, May 2019 ISSN No.: 2279-543X UGC Journal No.: 64650 The theory of problem solving based on soft computing proposed by Gupta and Kulkarni [2]. The figure on the left shows the traditional approach to hard calculations. There, an accurate model of the plant is available for research, and traditional mathematical methods are used to solve the problem. The correct figure shows that only the facility's inference model is available, and the solution shows a flexible computing approach that relies on the inference approximation method. This article gives an overview of flexible computing techniques and general techniques for solving complex problems using flexible computing techniques such as fuzzy logic, neural networks, genetic algorithms, and expert systems. 2. Fuzzy Logic The concept of topological logic was introduced by Zadeh [3] as a way to express human knowledge and is essentially mysterious. Figure 1 shows the basic configuration of the fuzzy logic system. The fuzzy interface converts net input values into fuzzy language values. A blur is still necessary in the fuzzy logic system since the input values of the current sensor are always numeric. The injection engine is based on fuzzy inputs and fuzzy rules and produces fuzzy outputs. The basis of the fuzzy rules is the form of the rule "IF-THEN" that contains language variables. The final processing element of the fuzzy logic system is the definition of generating a net output action. The main advantage of fuzzy logic is to provide a practical way to design nonlinear control systems that are difficult to design and stabilize. Use the traditional method. The structure of the fuzzy logic is shown in Figure 1. Fig. 1 Architecture of Fuzzy Logic 3. Artificial neural networks Neural networks (NN) and support vector machines (SVM) are part of machine learning, and perceptrons are part of neural networks. Artificial Neural Network (ANN) is one of the fastest growing research areas, attracting researchers from various engineering fields such as electronics, control engineering and engineering. software. ANN is an information processing system inspired by nervous system and brain functions. ANN is typically configured for specific applications such as pattern recognition, data validation, image processing, inventory forecasting, weather forecasting, image compression, 757 International Journal of Scientific Research and Review Volume 07, Issue 05, May 2019 ISSN No.: 2279-543X UGC Journal No.: 64650 security and security. I am ready. Neural networks aim to bring traditional computers closer to the work of the human brain. ANNS works best when the input-output relationship is highly non-linear. ANN is best for solving problems that do not follow a particular algorithm or set of rules to solve the problem. A neural network is a large network of interconnected elements inspired by human neurons. Each neuron performs a small operation, and the whole operation is a weighted sum of these operations. The neural network must be formed so that the set of known inputs produce the desired output. Training is usually done by sending an education model to the network, and some networks can modify their weighting functions according to predefined education rules. Learning can be monitored or not required. During the investigation he is trained in network monitoring by providing corresponding input and output models. That is, the result is known for a particular entry. In unreadable learning, the output of the network is shaped to respond to the input model. There are some advantages and disadvantages of the neural network:      ANNs are not universal devices for solving problems because there is no method of training and auditing ANNs. The outcome of ANN depends on the correctness of the existing data. Complex ANN systems may require excessive training ANN can handle incomplete data sets ANN succeed in forecasting and forecasting applications ANN basically consists of three layers. Entry, hidden layer, and exit. Each layer can have multiple nodes. Backpropagation algorithms are used as the method of network formation in most ANN networks. Here, the output of the neural network is evaluated with respect to the desired output. If the results are not expected, the weights between layers are corrected and this process is repeated until very small errors persist. The basic structure of ANN is shown in Figure 2.. Fig. 2 Basic architecture of three layer ANN 758 International Journal of Scientific Research and Review Volume 07, Issue 05, May 2019 ISSN No.: 2279-543X UGC Journal No.: 64650 3. Genetic algorithms Genetic and differential algorithms are part of evolutionary algorithms, which are also part of evolutionary computation (EC). Genetic algorithms are mainly used to solve various adaptation problems in real applications. The basic idea of genetic algorithms is to naturally reproduce natural selection in order to make a suitable choice for the application. Genetic algorithms are basically models of machine learning inspired by the process of development in nature. Solutions found in engineering applications [4]. Genetic algorithms can be used to find complex search problems. For example, they can search The best and least expensive project is to find the best combination between different models and components. Genetic algorithms are used in many areas such as climatology, biomedical engineering, coding, control, game theory, electronic design and automated manufacturing and design. The basic process of genetic algorithm is:     Initialize. The initial population is randomly created. Assessment-Each member of the population is assessed to assess how well an individual's physical condition meets their desired needs. Selection-Only those that meet your needs are selected. Crossovers that combine the strengths of existing systems to create new ones. In the end, we plan to create individuals closer to the desired needs.. The procedure is frequent from the second phase until it rangesanconclusionsituation. 4. Adaptive Neuro Fuzzy Inference System (ANFIS) ANFIS is collection of ANN methods and fuzzy system was established by Jung [4]. Fuzzy infection systems are classified into three types based on results of fuzzy rules: Tsukamoto System, Sugano System, and Damage System. Researcher commonly use the fuzzy models of Sugano and are therefore used in this study to predict soil dispersion. In the Sugano type estimation, two customizable methods, namely backpropagation and hybridization (from backpropagations and at least sections), use the membership function (ie, triph, galmoph, trichum, gausff, pymph, gauss 2mf, sigmoff) made to update. And Dsigmf) [5]. In the first order, in the Sugano system, if there are two inputs of x and y f 1 and f2 is its output, two fuzzy IF / THEN rules can be displayed as follows: Rule 1. If x is A1 and y is B1 then f1 = p1x + q1y + r1 and Rule 2. If x is A2 and y is B2 then f2 = p2x + q2y + r2 where x and y are the net inputs of the node i, Ai and Bi are the fuzzy sets of the antecedent, fi and f2is the output in the fuzzy region specified by the fuzzy rule; and pi, qi and ri are the 759 International Journal of Scientific Research and Review Volume 07, Issue 05, May 2019 ISSN No.: 2279-543X UGC Journal No.: 64650 design parameters determined during the training process. The basic structure of the ANFIS model is illustrated in Figure 3. Fig.3 Structure of ANFIS 5. Gene Expression Programming (GEP) GEP is an extension of Genetic Programming (GP) and Genetic Algorithm (GA). It was invented by Ferreira in 1999. It takes a population of individuals, selects them according to their physical form, and uses one or more ethers of genetic manipulation. Current genetic modification [5-6]. In recent years, GEP, GA, and GP models are widely used in data mining applications. GEP models are superior to GA and GP models. The advantages of the GAAP and GP models of the GAP model are based on its structure. The first step in developing a GEP model is to select a formatting function that can be created based on the error between the projection and the model comment. The next step is to select the set of ports T (variables or constants used in the problem) and select the function. (?), Subtraction (-), division (/), and multiplication (*). The third step in making a chromosome is to design a colorful architecture that includes the number of genes, the size of the head, and the joining function. The fourth step is to select the sub-ET link function. In the fifth stage, implementation of genetic operators (mutation, transpiration, inversion, cross hybridization / recombination and gene crossover) for chromosomal modification is performed.. 6. Particle Swarm Optimization Particle swarm optimization (PSO) is based on the idea of swarm intelligence to find the best solution at a specific research location. PSO is initiated by a random group of individuals and generations are optimized by updating. In each generation, each person is updated with the 760 International Journal of Scientific Research and Review Volume 07, Issue 05, May 2019 ISSN No.: 2279-543X UGC Journal No.: 64650 two best positions. The first value so far is the best state of the person called Pbest. The second best solution following this person is the overall best solution of Gbest. This is the best position in a complete survey of the whole population from past to present. In other words, each person in the population updates the position based on the best position and the state of the swarm. The search uses the root mean square error (RMSE) as the objective function. Low values indicate that the accuracy of the RMSE model is high. Regardless of whether the criteria are met, the PSO algorithm evaluates the objective function to meet the criteria. 7. Support Vector Machines Support Vector Machine (SVM) is a newly developed method. Support vector machines are derived from statistical teaching principles and can be used to determine the time taken from training using previous data. The main strength of SVM is the use of Structural Risk Minimization (SRM) rather than Empirical Risk Minimization (ERM). SVM is superior to traditional educational algorithms such as artificial neural networks (ANN) because it solves quadratic optimization problems that guarantee global compatibility. The resulting model is sparse and is not characterized by a "curse of dimension". SVM models are better suited for predicting soil moisture than ANN models. [7] Fig. 4 Flow diagram of SVM 8. Ant Colony Optimization Ant colony optimization (ACO) algorithm, as shown in Figure 5, is a potential technique for solving computational problems that can be reduced to find an appropriate path through the graph. In organic ant, pheromone-based communication is the most commonly used model. The combination of artificial ants and local search algorithms has become a suitable method for many personalization tasks, including certain types of graphics such as car routes and internet routes. In this area, the practices dedicated to special companies like artificial ants and Antôtima have been inspired by many business applications, for growing activities. 761 International Journal of Scientific Research and Review Volume 07, Issue 05, May 2019 ISSN No.: 2279-543X UGC Journal No.: 64650 Fig. 5 Ant Behaviour For example, Ant Colony Optimization [9] is a section of the optimization algorithm created by ant colony work. Artificial “ants” (like simulation agents), parameters representing all possible solutions, find the best solution while traversing space. Real ants gather pheromone and search for resources while searching for their own environment. Since the fake "ants" record the quality of their situation and their solutions, more simulations will get more solutions for better simulation. A variation of this approach is a bee algorithm that is suitable for beekeeping and other social insects that are models of bee beekeeping. This algorithm is a member of the Ant Colony Algorithm family of swarm intelligence and is a meta-heuristic optimization. 9. Probabilistic Reasoning Probabilistic reasoning is that a doctor examines a patient based on his / her medical history, test results, and prescribed symptoms, and the odds of 90% of the patient are satisfied, but some tests or results are unpredictable It may be missing. In other words, such an event is a probability. Probability provides a way to manage uncertainty due to laziness and ignorance. Probability-based reasoning is an understanding from the knowledge that there is uncertainty in this event. Some of the sources of uncertainty are personally-obtained information, experimental errors, random events that occur during major events and equipment failures. 10. Bayesian Network Bayesian networks, arrays of arrays, reliable networks, decision networks, Bayan models (ian), or similar guided models represent a series of variables and their dependencies. Graphic Bayesian Networks (DAGs) are ideal for considering such events, and predict the possibility that any of a number of known reasons may be a factor. For example, a Bayesian network can represent a potential relationship between disease and symptoms. By looking at the symptoms, the network can be used to calculate the possibility of the presence of various diseases. 762 International Journal of Scientific Research and Review Volume 07, Issue 05, May 2019 ISSN No.: 2279-543X UGC Journal No.: 64650 Eligible algorithms can perform intense learning in Bayesian networks. Bayesian networks that model variable sequences (such as speech signals and protein sequences) are called dynamic bisexual networks. A generalization of Bayesian networks that can express and solve the decision problem in uncertainty is called an effect diagram. 11. Conclusions and Future The above soft computing techniques are becoming more and more important as the capabilities of computer processing equipment increase and their cost goes down. Intelligent systems need to use complex algorithms to make complex decisions and to select the best result from many possibilities. This requires fast processing power and large storage space available at many research centers, universities and technical schools in recent years at very low cost. With the power and awareness of the Internet of Things concept, the need to build smart systems using flexible computer technology is more important than ever. Today, most soft computing applications can be effectively controlled by inexpensive but ultra-fast microcontrollers. We already see many appliances needed for daily life, such as washing machines, stoves, fuzzy logic in refrigerators, artificial neural networks, expert systems and more. Many industrial and commercial applications for soft computing are also in daily use and are expected to grow in the next 10 years. REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. Zadeh LA. Fuzzy logic, neural networks and soft computing. One-page course announcement of CS 294-4. Spring 1993. University of California at Berkeley; Nov. 1992. Gupta P, Kulkarni N. An Introduction of Soft Computing Approach over Hard Computing. International Journal of latest Trends 1993 Zadeh LA. Fuzzy logic, neural networks, and soft computing. Communications of the ACM 1994; vol. 37. no. 3. pp. 77-84. Mitchell M (1996) An introduction to genetic algorithms. MIT press, London Ferreira C (2001) Gene expression programming: a new adaptive algorithm for solving problems. Complex Syst J 13(2):87–129 Ferreira C (2001) Gene expression programming in problem solving. In: Invited tutorial of the 6th online world conference on soft computing in industrial applications, September 10–24 Pal, M. and P.M. Mather, 2003. Support Vector Classifiers for Land Cover Classification. Map India, Image Processing and interpretation Monmarché Nicolas, Guinand Frédéric and Siarry Patrick (2010). Artificial Ants. Wiley-ISTE. ISBN 978-1-84821-194-0. Dorigo, Gambardella, M, L.M. (1997). "Learning Approach to the Traveling Salesman Problem". IEEE Transactions on Evolutionary Computation, 1 (1): 214 Ant Colony Optimization by Marco Dorigo and Thomas Stützle, MIT Press, 2004. ISBN 0-262-04219-3 Friedman N, Geiger D, Goldszmidt M (November 1997). "Bayesian Network Classifiers". Machine Learning. 29 (2–3): 131–163. doi:10.1023/A:1007465528199 763