KBANN

KBANN: Knowledge-Based Artificial Neural Network

KBANN is one of the first work that allow the injection of symbolic rules into a neural network. Back in 1994, G. G. Towell introduced the concept of KBANN in this famous paper. In his work, he proposed an algorithm to build a NN from propositional logic rules.

About the rules

The supported rules of Towell’s algorithm are quite limited. Rules must be propositional logic rules and the handles type of data is categorical data. The rules must be classification rules (no other types of ML tasks are supported).

Example

Towell validated his method on a biological domain: the classification of splice-junction gene sequences.