C ++ artificial intelligence algorithm series 1 (basic knowledge of neural network algorithms)

Neural Network It is a collection of neurons, and neurons have the following behavior in the biological perspective:
(1) being restrained or excited;
(2) explosive growth and platforms;
(3) can generate a reverse attack after suppression;
(4) adaptability.
It has the following features:
1. Activation and inhibition of neurons
The activation and inhibition of a neuron are characterized by different potentials between the inside and outside of the cell membrane.
In the inhibition state, there is a positive potential difference between the inside and outside the cell membrane, which is about-50-mV.
In the excited state, the opposite potential difference between the inner and the outer is generated, and the electric pulse is about 60.
The pulse width of excited cells is generally about 1 ms.
2. Information Transmission and threshold characteristics of neurons
When neurons receive information from other neurons, the membrane potential changes continuously at the beginning.
When the change in membrane potential exceeds a certain value, a sudden increase pulse is generated, which is then transmitted along the axon.
The feature called the threshold feature that generates pulse transfer when the membrane potential of neurons reaches a threshold value.
One-way transmission (that is, it can only transmit the DC or cell body of the previous first-level neuron to the back-level neuron, not vice versa .)
And latency (information is transmitted through the SYN, which usually produces a latency of 0.5-1 ms .)
3. Information synthesis of neurons
Neurons have the comprehensive characteristics of time and space for information from other neurons. In terms of neural network structure, a large number of different neurons
The peripheral of an Eton can be distributed to the same neuron to form a large number of syncs. Neurons released from different syncs
It can affect the changes of membrane potential of the same neuron. Therefore, on the DC, neurons can
The input information is integrated. This is the comprehensive spatial feature of neurons for information. For information from the same SYN, neural
The metadata can be integrated with the Information transmitted at different times. Therefore, neurons have a time-based synthesis feature for information.
4. D/A and A/D properties of neurons
The information transmitted from the neurons is the same amplitude, constant width, and encoded discrete electrical pulse signal, so it is a number. However
The release of neurotransmitter and the change of membrane potential in the DC are continuous. Therefore, it indicates that the SYN has the D/A function. In God
When the DC membrane potential of the element is higher than a threshold value, it becomes an electrical pulse that is transmitted by the axon. Therefore, this process is described
Neurons have A/D function. In this way, when information is transmitted through a neuron, the neuron performs the D/A and A/D conversion processes on the information.

Attachment: the entrance to receiving information from other neurons;
The axon refers to the transfer of neuron excited information to the outlet of other neurons;
A syn is a structure in which a neuron is associated with another neuron and transmits information.