Analysis of neural network computing power

In this paper, the nonlinear dynamic properties of neural networks are studied. The dynamic system theory, nonlinear programming theory and statistical theory are mainly used to analyze the evolution process of neural networks and the properties of attractors, explore the cooperative behavior and collective computing function of neural networks, and understand the neural information processing mechanism. In order to explore the possibility of neural network processing information in terms of integrity and fuzziness, the concept and method of chaos theory will play a role. Chaos is a very difficult mathematical concept to define accurately. Generally speaking, "chaos" refers to the nondeterministic behavior in the dynamic system described by deterministic equations, or deterministic randomness“ "Certainty" is because it is caused by internal reasons rather than external noise or interference, while "randomness" refers to its irregular and unpredictable behavior, which can only be described by statistical methods. The main characteristic of chaotic dynamic system is the sensitive dependence of its state on initial conditions, and chaos reflects its inherent randomness. Chaos theory refers to the basic theory, concept and method of describing the nonlinear dynamic system with chaotic behavior. It understands the complex behavior of dynamic system as its own internal structured behavior in the process of material, energy and information exchange with the outside world, rather than external and accidental behavior. Chaotic state is a kind of steady state. The stationary states of chaotic dynamical systems include: stationary, stationary, periodic, quasi synchronous and chaotic solutions. Chaotic trajectories are the result of the combination of global stability and local instability, which are called singular attractors. A singular attractor has the following characteristics: (1) a singular attractor is an attractor, but it is neither a fixed point nor a periodic solution 2) Singular attractors are indivisible, that is, they cannot be divided into two or more attractors 3) It is very sensitive to the initial value, different initial values will lead to very different behavior

After decades of development, neural network theory has achieved extensive success in many research fields, such as pattern recognition, automatic control, signal processing, assistant decision-making, artificial intelligence and so on. The following introces the application status of neural network in some fields. In dealing with many problems, the information source is not complete, but also contains false, sometimes the decision rules are contradictory, sometimes there is no rules to follow, which brings great difficulties to the traditional way of information processing, but neural network can deal with these problems well, and give reasonable identification and judgment
1. Information processing
the problems to be solved in modern information processing are very complex. Artificial neural network can imitate or replace the functions related to human thinking, realize automatic diagnosis and problem solving, and solve the problems that traditional methods cannot or are difficult to solve. Artificial neural network system has high fault tolerance, robustness and self-organization, even if the connecting line is damaged to a high degree, it can still be in the optimal working state, which is widely used in military system electronic equipment. The existing intelligent information system includes intelligent instrument, automatic tracking and monitoring instrument system, automatic control and guidance system, automatic fault diagnosis and alarm system, etc
2. Pattern recognition
pattern recognition is a process of describing, identifying, classifying and explaining things or phenomena by processing and analyzing various forms of information representing things or phenomena. The technology is based on Bayesian probability theory and Shennong's information theory, and its information processing process is closer to the logical thinking process of human brain. Now there are two basic pattern recognition methods, statistical pattern recognition method and structural pattern recognition method. Artificial neural network is a common method in pattern recognition. In recent years, the pattern recognition method of artificial neural network has graally replaced the traditional pattern recognition method. After years of research and development, pattern recognition has become a relatively advanced technology, which is widely used in character recognition, speech recognition, fingerprint recognition, remote sensing image recognition, face recognition, handwritten character recognition, instrial fault detection, precision guidance and other aspects. Due to the complexity and unpredictability of human body and disease, the detection and signal expression of biological signals and information, the analysis and decision-making of acquired data and information, and many other aspects have very complex non-linear relations, which are suitable for the application of artificial neural network. The current research almost involves all aspects from basic medicine to clinical medicine, mainly used in the detection and automatic analysis of biological signals, medical expert system and so on< Detection and analysis of biological signals
most medical detection devices output data in the form of continuous waveforms, which are the basis of diagnosis. Artificial neural network is an adaptive dynamic system composed of a large number of simple processing units. It has the functions of massive parallelism, distributed storage, self-organization of adaptive learning and so on. It can be used to solve the problems that are difficult or impossible to be solved by conventional methods in biomedical signal analysis and processing. The application of neural network in biomedical signal detection and processing mainly focuses on the analysis of EEG signal, the extraction of auditory evoked potential signal, the recognition of EMG and gastrointestinal electric signal, the compression of ECG signal, the recognition and processing of medical image, etc
2. Medical expert system
the traditional expert system is to store the experience and knowledge of experts in the form of rules in the computer, establish a knowledge base, and carry out medical diagnosis by logical reasoning. But in practical application, with the increase of database scale, it will lead to knowledge "explosion", and there are "bottleneck" problems in the way of knowledge acquisition, resulting in low work efficiency. Neural network based on non-linear parallel processing points out a new development direction for the research of expert system, solves the above problems of expert system, and improves the reasoning, self-organization and self-learning ability of knowledge. Therefore, neural network has been widely used and developed in medical expert system. In the research of anesthesia and critical care medicine and other related fields, it involves the analysis and prediction of multiple physiological variables. In clinical data, there are some relations and phenomena that have not been found or have no exact evidence. Signal processing, automatic discrimination and detection of interference signals, and prediction of various clinical conditions can be applied to artificial neural network technology. 1. Market price forecast
the analysis of commodity price changes can be summarized as a comprehensive analysis of many factors affecting the relationship between market supply and demand. Because of the inherent limitations of traditional statistical economics, it is difficult to make a scientific prediction of price changes, and the artificial neural network is easy to deal with incomplete, fuzzy and uncertain or irregular data, so using artificial neural network for price prediction has advantages that traditional methods cannot compare. Starting from the determination mechanism of market price, according to the number of households, per capita disposable income, loan interest rate, urbanization level and other complex and changeable factors that affect the commodity price, a more accurate and reliable model is established. The model can scientifically predict the trend of commodity prices and get accurate and objective evaluation results
2. Risk assessment
risk refers to the possibility of economic or financial loss, natural damage or damage caused by the uncertainty in the process of engaging in a specific activity. The best way to prevent risks is to make scientific prediction and assessment of risks in advance. The application of artificial neural network is to construct the structure and algorithm of the credit risk model suitable for the actual situation according to the actual risk sources, get the risk evaluation coefficient, and then determine the solution of the actual problem. Using this model for empirical analysis can make up for the lack of subjective evaluation and achieve satisfactory results. From the beginning of the formation of neural network model, it is closely related to psychology. Neural network is abstracted from the information processing function of neurons, and the training of neural network reflects the cognitive process of sensation, memory and learning. Through continuous research, people change the structure model and learning rules of artificial neural network, and explore the cognitive function of neural network from different angles, which lays a solid foundation for its research in psychology. In recent years, artificial neural network model has become an indispensable tool to explore the mechanism of advanced psychological processes such as social cognition, memory and learning. Artificial neural network model can also be used to study the cognitive impairment of patients with brain injury, which challenges the traditional cognitive positioning mechanism
although the artificial neural network has made some progress, there are still many defects, such as: the application is not wide enough, the results are not accurate enough; The training speed of existing model algorithms is not high enough; The integration degree of the algorithm is not high enough; At the same time, we hope to find a new breakthrough in theory and establish a new general model and algorithm. Further research on biological neuron system is needed to enrich people's understanding of human brain