Kanerva's approach has been championed by Jeff Hawkins' Numenta and Redwood Neuroscience Institute. Numenta uses the terminology Sparse Distributed Representations (SDR).

While the theoretical basis seems sound, a practical implementation comparable to the SOTA is lacking.

Ha, I wrote most of this article. Also check out https://en.wikipedia.org/wiki/Holographic_associative_memory but it needs some work

Double Ha! I just posted that too, in a comment last week...

Discussion: Algebra and the Lambda Calculus (1993) https://news.ycombinator.com/item?id=18296384

Is holographic associative memory one of your areas of research too?

it's the closest we can get to modeling something like phase precession or phase resetting in grid cells http://www.scholarpedia.org/article/Grid_cells

Any good papers/code examples that discuss data model design and implementation details?

Links aside, it seems there's another out there with a kindred username!

Also, great write-up. That article is awesome!

I don't keep around many items from my collection of my mid-90s AI books, but for some reason Kanerva's Sparse Distributed Memory book always managed to make the cut whenever I go in to cull things. There is something about the idea that seems appealing and I keep waiting for some excuse to see if any of the ideas can be melded with more recent approaches to NN memory...