A New Type of Architecture for Neural Networks with Multi-connected Weights in Classification Problems

Studies have shown that the interaction of biological neurons is based on neurotransmitters, which transmit signals between neurons, and one neuron sends information to another neuron by releasing different neurotransmitters that play different roles. Based on this biological approach, a new model of neural networks is proposed by increasing the amount of connection weights between two neurons, that is, based on the assumption that there are several connections between each connection from a biological point of view. The sum of the binding weights represents the sum of the neurotransmitter category, and different components of the weights correspond to different neurotransmitters. Inputs and outputs are determined heuristically for each link in the proposed model so that these neurotransmitters compete appropriately. From a biological point of view, the proposed neural network models can be obtained as mathematical models that are closer to biological neural networks. From the point of view of the structure of new models, the fact that the activation of each hidden neuron is based on several filters can improve the interpretation of the features learned by the neural network.