The Memory Tesseract: Distributed MINERVA and the Unification of Memory

We prove that MINERVA 2, a widely-used model of biological long-term memory, is mathematically equivalent to an auto-associative memory implemented as a fourth order tensor. We further propose an alternative implementation of MINERVA 2 as a holographic lateral inhibition network. Our work clarifies the relationship between MINERVA 2 and other memory models, and shows that MINERVA 2 and derivative models can be neurally implemented and scaled-up to longterm learning tasks.