Re: Proof that neural nets work

Hi William,

You got me there, a neural net is a neural net. The only real
difference could be if a person made a mistake programming it up, used
the wrong transition function etc.

The fact that GoldenGem is proven to actually work, means something
but admittedly not a lot.

Also the fact it runs once you enter the tickers, no messing around.

Also the fact that the 'backprop' algorithm is the right one. Again
admittedly there is only one right algorithm, but a lot of the people
(in engineering) who write about and use neural nets don't always get
the partial differential equations right.

The GoldenGem calculator is just an industry standard, I'm trying
to bring it into the 21'st century. No claim or originality on my part,
but I really understand all the main issues and the performance proves
itself.

John Moody.

.

Relevant Pages

  • Re: Associative memory for navigation
    ... visual layout of a building -- using some kind of neural net. ... That algorithm looks like it could be useful. ... whatever data structure is used by this SIFT algorithm, ... So I'd send raw image data as input to the ANN, ...
    (comp.robotics.misc)
  • Re: string class with method approximatelyEquals(String s)
    ... >The hardest part of software development is not in telling ... Then a neural net ... algorithm works out a scheme to put your intuition into algorithmic ... Coaching, problem solving, economical contract programming. ...
    (comp.lang.java.help)
  • Re: Associative memory for navigation
    ... visual layout of a building -- using some kind of neural net. ... just pass the raw image data for the images to the algorithm each time ... robot, and I used SIFT. ...
    (comp.robotics.misc)
  • New Training Algorithm for Feed Forward Neural Nets
    ... I'm currently working on a new Training Algorithm for Feed Forward Neural ... convert the neural net into a SAT instance ... It maybe that there is no exact solution so my hope is that MaxSat is ...
    (comp.theory)
  • Re: Making Java extinct!
    ... various awkward cases and the computer creates an interpolating ... algorithm to do something sensible for ... it still sort of works. ... some experience with implementing neural net based logic and i dont think they're extremely suitable for applications where exact results are mandatory and the behaviour completely predictable. ...
    (comp.lang.java.programmer)