Re: neurosymbolic approaches

Another related research area is on graph neural networks:

"A Comprehensive Survey on Graph Neural Networks”
 https://arxiv.org/pdf/1901.00596.pdf <https://arxiv.org/pdf/1901.00596.pdf>

This is more closely linked to deep learning, and shows that there are a variety of approaches for work on graph data involving different and unconnected research communities.

My feeling is that as AGI is realised, a lot of what we do now will seem quite odd in retrospect. We will have been seen to be working on asking rather futile questions.

> On 17 Feb 2020, at 05:04, Paola Di Maio <paoladimaio10@gmail.com> wrote:
> 
> Dave, thanks
> No I did not have that reference, did not come up in searches, shall check it out
> any chance you may want to enter it in the zotero library
> thank you!!
> PDM
> 
> On Sat, Feb 15, 2020 at 6:59 PM Dave Raggett <dsr@w3.org <mailto:dsr@w3.org>> wrote:
> Hi Paola,
> 
>> On 15 Feb 2020, at 01:17, Paola Di Maio <paola.dimaio@gmail.com <mailto:paola.dimaio@gmail.com>> wrote:
>> 
>> I am working on a Neurosymbolic approaches for AI KR.
> 
> I hope that your paper will cite Chris Eliasmith’s computational neuroscience team at the University of Waterloo. They have worked extensively on biologically accurate models describing how symbolic information can be expressed and manipulated by neural networks.
> 
>  https://uwaterloo.ca/systems-design-engineering/profile/celiasmi <https://uwaterloo.ca/systems-design-engineering/profile/celiasmi>
> 
> "Higher-level cognitive functions in biological systems are made possible by Semantic Pointers. Semantic Pointers are neural representations that carry partial semantic content and are composable into the representational structures necessary to support complex cognition."
> 
> Semantic pointers are vectors in n-dimensional spaces corresponding to concurrent patterns of firing across bundles of nerve fibres. It is essentially the same idea as a chunk, i.e. a collection of properties, whose values identify other chunks. Using addition and circular convolution, multiple semantic pointers can be stored and retrieved within a single vector.  Mathematical details are available at:
> 
>  https://www.nengo.ai/nengo-spa/user-guide/spa-intro.html <https://www.nengo.ai/nengo-spa/user-guide/spa-intro.html>
> 
> There is also a downloadable python package for testing and deploying neural networks, see: https://www.nengo.ai/ <https://www.nengo.ai/>
> 
> Best regards,
> 
> Dave Raggett <dsr@w3.org <mailto:dsr@w3.org>> http://www.w3.org/People/Raggett <http://www.w3.org/People/Raggett>
> W3C Data Activity Lead & W3C champion for the Web of things 
> 
> 
> 

Dave Raggett <dsr@w3.org> http://www.w3.org/People/Raggett
W3C Data Activity Lead & W3C champion for the Web of things