Webneurons. A new way to program computers.

Current programming languages are usually full of symbols and jargon and it can be very difficult to follow the flow of the program and to trace variables etc. These problems are mostly due to the historical assumption that programming should be scripted like a book is written.

The average person can make little headway against these obstacles and many steer clear of programming. It is partly on their behalf that a simpler solution is suggested.

The greatest computer is the brain. Presumed to have evolved over billions of years this demonstrates that the best programming may not be scripted, function or object based, but rather network based. Those of you with a little knowledge of brain anatomy must notice the striking similarity to the World Wide Web taking each brain neuron as a web page and brain axons as hyperlinks. The big difference is that the Web is largely a browsable corpse with no automatic data transfer between pages. Add this missing ingredient and the Web should come to life, and with hundreds of new applications.

The basic idea is that mini, generated web pages (webneurons) take the place of conventional programming functions and objects, and that active hyperlinks between these pages take the place of program flow, e.g looping and function calls.

Imagine browsing a network of webpages, moving data between pages as you go, and also performing a simple operation on the data held in each page. Now imagine that this process was done automatically, then that is how webneurons work. Web pages and hyperlinks now become active and any page could automatically stimulate the activity of many other pages which could stimulate others and so on. Cascades, looping, if/then, recursion etc can all be handled.

The whole system can be edited by using browsing and with user friendly indexes, clickboards, page titles, etc in fact all the tools of the present day browsable Web, rather than wading through lines and lines of undecipherable script. Instead of being a separate item, programming itself could finally be correctly integrated with the Web. Webneuron networks have some similarities to neural networks but without any restrictions on node linking. For example there can be as many layers as you like.

Webneurons can be used in modelling N dimensional applications. This is possible because you can just keep adding links to each webneuron in the network. So a one dimensional system can be modelled by a single line of webneurons with a link to each nearest neighbour. A 2D system involves 4 links per webneuron and typically this could model a simple database with each webneuron representing one field and each line representing one record. 3D is handled by 6 links and typical applications might include graphics. 4D has 8 links and so on. Following an original post to comp.ai and comp.ai.neural-nets on 21 May 1996 there is now an immediately runnable demonstration program in javascript at:-

http://webneurons.com

This has been available for some time but has only just been placed online. It is not perfected but just gives a little idea perhaps of what might be possible.

The original post is archived at:-

http://plexos.com/Webneurons/564.htm

where there are many other pages on the subject.

This software has been licensed as GNU freeware and is also downloadable as part of a package at:-

http://plexos.com/master11.zip

Any proposals to develop the software further are welcomed.

--

John Marshall
Neuroweb Ltd
http://www.plexos.com