I think neuron networks are inspired by the biological neuron structure. As we all know, information is transmitted in the form of an electric signal through each neuron. The outside stimuli will be transferred to a neurotransmitter that will be accepted by the brain eventually. Human beings will do reaction according to the influence they receive from the neurotransmitter. Neuron networks, similarly, will receive the input data and process calculation. Since the function of activation function, the input will be transferred to the “neurotransmitter” of neuron networks—the result of the activation function(0 or 1, 0 or the sum of input, etc. Depends on the activation function). These “neurotransmitters” will be pass from the hidden layer to the next neuron of other layers. After going through all the layers of the neuron network structure, the programming will reach to the final output, and make reaction according to the output. Hence, we can conclude that neuron networks mimic the operating principle of the human body’s neuron structure and brains.
However, there are also differences between neuron networks and neuron structure of the human body. Since programming makes prediction instead of responding accordingly as the neuron system of humans do, we need to use abundant data to train the programming model so that the accuracy of neuron networks will be improved. What’s more, the direction of the correcting process is from the output layer to the input layer, which is a mechanism that the human neuron system doesn’t have. From the output layer to the input layer, each layer will correct the mistakes it makes. But in the human neuron system, the neuron signal’s conduction has its directivity. There is no way that neurons can reverse the direction of the signal’s transmit and pass the signal back to the last neuron.
Deep neural networks have something in common with the human neuron system, but we cannot say something like “the deep neural networks are completely simulating”. The way both two solve problems is the same.
Leave a Reply