What can we learn about efficient computation from looking at small biological neural networks, like those in insects? We need a non-procedural, non-von Neumann understanding of computation to understand biological neural processing. Designing for FPGAs might be a better software background to have when approaching spiking neural networks. We could adopt clockwork computation as a design paradigm. In clockwork, everything happens all the time. There is no concept of a beginning or end to a computation — computation is a continual processing of information. Should we imagine that computation in biological neural networks involves reaching some final result or fixed equilibrium point? Might it not be more likely that on-going activity represents a continual “best guess” result? We can adapt these concepts to low-power, close-to-the-sensor processing

IX

When the blackbird flew out of sight,
It marked the edge
Of one of many circles.

Thirteen Ways of Looking at a Blackbird — Wallace Stevens.

Networking

twitter:

View profile

github:

View code

LinkedIn:

View profile

email:

Erdos number:

4

Colophon

This website was written by hand, using PHP, HTML5 semantic markup and CSS2.1 / CSS3. The layout and styles are my own, but with a heavy nod to Oliver Reichenstein's text as user interface manifesto.

The text on this site is set in Sorts Mill Goudy by Barry Schwartz. This font is a revival of Goudy old style, an Aldine classic serif typeface originally created by Frederic W. Goudy for American Type Founders (ATF) in 1915.