Model Export for the Convolutional Coalesced Tsetlin Machine

This article describes how to export a model from a trained Convolutional Coalesced Tsetlin Machine (ConvCoTM). Specifically, the encoding of the literals, including the patch position information, and the exact sequence of the associated include/exclude bits in the software model, is described. The method was tested and verified in VHDL of a ConvCoTM inference-only hardware design.