Logical Approach to Interpreting Neural Network Solutions

Neural network decision interpretation is the process of understanding how the neural network makes decisions based on input data and what factors have the greatest influence on those decisions. The importance of interpreting neural network decisions lies in the fact that it can help to understand which factors are most significant for decision making, and help improve network performance, as well as find hidden patterns in the subject area. The paper proposes an approach using Boolean integro-differential calculus. Possible logical relationships between input data and decision results are considered. Next, a function is built that allows you to analyze in detail the decision domain of the neural network, find the most important features and hidden patterns in the data. This is especially useful for solving problems.