Explainable neural net architecture for multidimensional data

A computer implemented method for providing a convolutional explainable neural network, comprising:

• receiving, from a black-box convolutional neural network, based on an input provided to the black-box convolutional neural network, a set of output data provided as an output of the black-box convolutional neural network;
• identifying a relationship between the input and the output, generating at least one predicted output based on the relationship, and generating a set of data based on the at least one predicted output;
• convoluting the set of data into a plurality of features, inputting the plurality of features into an input layer, and partitioning the input into one or more partitions based on the plurality of features;
• constructing one or more local models associated with the one or more partitions, wherein constructing the one or more local models based on the one or more partitions comprises:
• aggregating one or more of the partitions;
• applying one or more transformation functions to the partitions, providing a set of transformed features;
• combining the transformed features and identifying one or more coefficients corresponding to the transformed features;
• and
• forming the one or more local models based on the coefficients and the transformed features;
• forming rules based on the one or more local models;
• representing the rules in a symbolic logic format;
• forming an explainable model based on the rules and the one or more local models;
• applying the explainable model to a set of input features to obtain an output value;
• and
• outputting the output value and outputting an explanation of the output value, wherein the explanation accounts for at least one of the coefficients or the plurality of features.