Three main branches of Computing Intelligence and Their Applications

1. Summary

In the field of computer science, the most meaningful and challenging question is the simulation of human intelligence. Computing Intelligence 3 uses modern computing tools to simulate human intelligence mechanisms, theories and methods of information acquisition, processing (solving problems), and utilization for Life Evolution and artificial intelligence behaviors. It is the deepening and development of AI. If artificial intelligence is based on a knowledge base, Computing Intelligence is an information system that includes data, algorithms, and implementations based on models and features of step-by-step, parallel, and bionic computing. The former emphasizes the form and representation of rules, the latter emphasizes the establishment and composition of models, the former relies on expert knowledge, and the latter emphasizes the self-organization, self-learning, and self-adaptation of systems.

Only three main branches of computing are: artificial neural networks, genetic algorithms, and fuzzy logic.

2. Keywords

Computing Intelligence; neural networks; Genetic Algorithms; Fuzzy Logic

3. Branch and Application of Computing Intelligence

3.1 Neural Networks

An artificial neural network system is a complex network system formed by a large number of simple processing units, namely the extensive connection of neurons. In artificial neural networks, computation is done by the flow of data in the network. In the process of data flow, each neuron receives the input data stream from the connected neuron and processes it, then, the results are transmitted to other neurons connected to them as output data streams. The topological structure of the network and the connection weight wi between neurons are determined by the corresponding learning algorithm. The algorithm constantly adjusts the connection weights between the network structure and neurons until the output required by the neural network is generated. Through this learning process, artificial neural networks can automatically acquire knowledge from the environment and store the knowledge in the network structure and connection weights.

Artificial Neural Networks have good self-learning, self-adaptation, and self-organizing capabilities, as well as parallel human scale, distributed information storage and processing, this makes it very suitable for dealing with incomplete and inaccurate information processing problems that require multiple factors at the same time. At present, artificial neural networks have been highly valued by academia and have been widely used in many fields. However, we can see that in the design process of neural networks, the settings of various parameters and the determination of the network structure are strongly empirical and there is no complete theory to follow, its scale is far from reaching the scale of tens of billions of neurons in the human brain. Moreover, artificial neural networks are based on brain models, and their research is limited by the results of Brain Science Research. Before we have a clear understanding of the brain's thinking patterns and cognitive processes, it is difficult to simulate the human brain.

3.2 Genetic Algorithm

During the evolution of billions of years, organisms in nature have formed an internal mechanism to optimize their own structures. They can constantly learn from the environment to adapt to the changing environment. For most organisms, this process is done through natural selection and sexual reproduction. Natural selection determines which individuals in the population can survive and multiply: Sexual Reproduction ensures the mixing and recombination of genes in future generations. Inspired by this natural evolution process, evolutionary computing starts by simulating the biological evolution process of the natural world and explores the laws of the development and evolution of some intelligent human behaviors at the genetic level, to solve the problem of how smart systems learn from the environment.

The theoretical basis of evolutionary computing is Darwin's theory of evolution. It is a new computing method formed by the combination of computer science and biological genetics. Evolutionary computing uses simple coding techniques to represent various complex structures. It also performs simple genetic operations and natural selection strategies for survival of the fittest to guide Learning and determine the search direction through a set of encoding techniques. Perform academic research by performing genetic operations such as population replication, hybridization, and mutation. Evolutionary Algorithms can search for multiple points in different regions of the solution space. They can find the global optimal solution with a high probability and are not easy to fall into the local optimal condition. The basic structure of evolutionary algorithms is as follows:

{An initial group is randomly generated to calculate the adaptive value of each individual in the group;

While does not meet the termination criteria do

{Apply genetic operations (replication, hybridization, mutation, etc.) to generate the Next Generation: Calculate the adaptive value of each individual in the population }}.

Evolutionary Algorithms do not know anything about solving the problem, but as long as they provide representation schemes, adaptive functions, genetic operators, control parameters, termination criteria, and so on, the algorithm can search for an unknown space in a way that does not rely on the problem itself, and finally find the solution to the problem. Evolutionary Algorithms are also simple, general, robust, and suitable for parallel processing, and intelligent features such as self-organization, self-adaptation, and self-learning, it has been successfully applied to complex problems that are difficult to solve using traditional methods. Evolutionary computing has shown its charm, especially in the areas of system identification, fault diagnosis, machine learning, and neural network design. However, as a new and interdisciplinary research subject, the theoretical research of evolutionary computing needs to be further improved, including basic theory, encoding mechanism, selection strategy of control parameters, and convergence analysis.

3.3 Fuzzy Logic

To express and deal with many inaccuracies and uncertainties in the real world, Zadeh proposed the fuzzy set theory in 1965. In a fuzzy set, the boundary of the set is not clear, and the membership of the set is not positive or negative. It uses a membership function to describe the intermediate transition of the symptom difference, this breaks through the absolute relationship between the classical set and the non-classical set. In a fuzzy set, each individual is assigned a value to indicate the extent to which the individual belongs to the set. This value reflects the similarity between the individual and the concept represented by the fuzzy set: the higher the degree of membership, the greater the degree of belonging to the set, and vice versa. The fuzzy system is based on fuzzy set theory and fuzzy logical reasoning. It attempts to simulate the ability of the human brain to express and solve inaccurate knowledge at a higher level. In a fuzzy system, knowledge is stored in the form of rules. It uses a set of fuzzy if then rules to describe the characteristics of objects and uses fuzzy logical reasoning to solve the uncertainty problem. Fuzzy systems are good at describing the use of knowledge in the subject field and have strong reasoning capabilities. Over the past 10 years, fuzzy systems have been widely used in expert systems, intelligent control, fault diagnosis, and other fields, and some exciting achievements have been made. However, further research is required on the issues of automatic extraction of Fuzzy Rules and Automatic Generation of membership functions.

 

4. Conclusion

As a new term, Computing Intelligence (CI) involves three fields: neural networks, fuzzy systems, and evolutionary computing. However, Computing Intelligence can integrate the three into an organic whole, so as to achieve complementary advantages. In this way, the integrated system will be more effective than a single technology and achieve greater results. The emerging Computing Intelligence expands the traditional computing model and intelligent theory, and is suitable for complex systems that cannot be accurately described using mathematical models. It has been widely used in the field of mechanical engineering, some results have been achieved. However, there are still many shortcomings in the theoretical and practical application of Computing Intelligence (such as the learning problem of neural networks), which restricts the practicality of Computing Intelligence, further efforts are yet to be made by researchers to improve the performance.

 

 

 

References

[1] Zhang Liming. Modeling and Application of Artificial Neural Networks [M]. Shanghai: Fudan University Press, 1993.

[2] NLP ordinary, et al. Artificial Neural Networks and fuzzy evolutionary computing [M]. Beijing: Tsinghua University Press.

[3] Chen Guoliang, et al. Genetic Algorithms and Applications [M]. Beijing: People's post and telecommunications Publishing House.