Understanding Foundation of Neural Network: Reasoning-based Approach

Background: Currently, neural network model needs understanding on what derives conclusion by execution. Neural network model imitates human intelligence but it has a limitation of cannot explaining what it is to construct the knowledge which it derives.

Object: By inducing universal framework from existing neural network models, common reasoning that regulates the intelligence that derives from the models would be introduced.

Method: Starting from minimizing model's bias, integrative method of understanding how neural network takes place in numerical ordinals and cardinals would be introduced. After that, explanation of intelligence by neural network would be described.

Result: Foundation of artificial neural network intelligence is an optimization with a grid of minimized error from ground truth.