Artificial neural network is the abstraction and Simulation of some basic characteristics of human brain or biological neural network . It provides a new idea for the research of many problems such as machine learning , At present, it has been in pattern recognition 、 Machine vision 、 associative memory 、 Auto-Control 、 signal processing 、 Soft measurement 、 Decision analysis 、 Intelligent Computing 、 Solving combinatorial optimization problems 、 Data mining has been successfully applied .
Focus on the most basic 、 Most typical 、 The most widely used BP Neural networks and Hopfield Neural network and its application in pattern recognition 、 associative memory 、 Soft measurement 、 Intelligent Computing 、 Application of combinatorial optimization problem solving .

neural network （neural networks,NN）

Biological neural networks ( natural neural network, NNN): By the central nervous system （ Brain and spinal cord ） And the peripheral nervous system （ Sensory nerves 、 Motor nerve, etc ） A complex neural network , The most important one is the brain nervous system .
Artificial neural network (artificial neural networks, ANN): Simulate the structure and function of human brain nervous system , An artificial network system composed of a large number of simple processing units and widely connected .
Neural network method ： Implicit knowledge representation

The structure of biological neurons

The structure of the human brain :
cortex （cortex）
midbrain （midbrain）
brainstem （brainstem）
cerebellum （cerebellum）

The human brain consists of more than 100 billion （1011 Billion － 1014 Billion ） A nerve cell （ Neuron ） A network of interwoven structures , The cerebral cortex is about 140 Billion neurons , The cerebellar cortex is about 1000 Billion neurons .
Neurons are about 1000 Types , Each neuron is about the same as 103－ 104 Connected to other neurons , Form an extremely complex and flexible neural network .

Biological neuron structure

Dendrites （ Input ）

Nerve cells use electricity - Chemical processes exchange signals
Working state ：
Excited state ： Cell membrane potential > Threshold of action potential → Nerve impulse
Suppression state ： Cell membrane potential < Threshold of action potential
Learning and forgetting ： Due to the plasticity of neuronal structure , Synaptic transmission can be enhanced and weakened .

Neuron mathematical model

Artificial neuron model

1943 year , Mccroach and Pitts proposed M－P Model .


The general model ：





Nonlinear excitation function （ Transfer function 、 Output transformation function ）

Neural network structure and working mode

Three factors that determine the performance of artificial neural networks ：

The characteristics of neurons .
The form of interconnection between neurons —— topology .
Learning rules to improve performance to adapt to the environment .

The structure of neural networks

（1） Feedforward （ Forward type ）

（2） Feedback

How neural networks work

Sync （ parallel ） The way ： At any time, all neurons in the neural network adjust the state at the same time .
asynchronous （ Serial ） The way ： Only one neuron adjusts at any one time , The state of other neurons remains unchanged .

Learning neural networks

Neural network method is a knowledge representation method and reasoning method .
Neural network knowledge representation is an implicit representation method .
1944 Nian Hebu （Hebb） A method to change the connection strength of neurons is proposed Hebb Learning rule .
Hebb Learning rule ： When neurons at both ends of a synapse are excited at the same time , Then the weight of the connection should be enhanced .
Welcome to join me for wechat exchange and discussion （ Please note csdn Add ）