Hyperdimensional Computing

Jul 23, 2021 Last reply: 4 years ago 11 Replies

Here is one of the fundamental papers on implementing associative memory (the analog parallel to content-addressable digital memory):



"Hyperdimensional Computing: An Introduction to Computing in Distributed Representation with High-Dimensional Random Vectors", Pentti Kanerva, Cogn Comput (2009) 1:139–159, DOI 10.1007/s12559-009-9009-8



.

formatting link



The part titled "What is the Dollar of Mexico?" on page 154 is what convinced me that Kanerva was on to something - he had explained how analogy works as a direct consequence of vector math over hyperdimensional spaces.


Reference [4] from 1988 is the founding article.



This is one reason I did not believe Penrose on Quantum Consciousness.


Kanerva presents Hypervectors:



.

formatting link


Joe Gwinn


This highlights the difference between meatware and software. Software is driven by imperatives: do this, then that (even if those are MIMD)

The brain is optimized for "like". It searches for something SIMILAR to the current input and then sorts out what to do (situationally) once it has that as a reference.

If you think of how you'd implement "like" in software (or hardware), you'd realize it is an incredibly costly operator. (and, even *defining* what you mean by "like" can be a problem in and of itself!)

What is done now is keep millions of humans in thrall, then throw models of "like" against the wall until one sticks.

Similar to this:

Join the Discussion

Have something to add? Share your thoughts — no account required.

Didn't find your answer?

Ask the community — no account required