Re: [xsl] Performance of XSLT in a neural network application

["Liam R. E. Quin liam@xxxxxxxxxxxxxxxx" <xsl-list-service@xxxxxxxxxxxxxxxxxxxxxx>]

On Sat, 2020-08-22 at 13:21 +0000, Roger L Costello costello@xxxxxxxxx
wrote:
> 
> Conclusion: XSLT is not a viable language for creating neural
> networks. 

We planted parsnip seeds in the garden and they didn't grow. But the
carrots grew. Conclusion, parsnips are not a viable food source?

What you have shown is that you have written a program that runs
slowly.

The first version of the text retrieval system i wrote in the 1980s
took ,ore than 12 hours to index a small test-set. The next version
took an hour. After several years of improving the system, it became
one of the two fastest text retrieval packages in existence and could
build the same index (actually with more information in it) in a matter
of seconds.

The question is how much it's worth to spend effort making a program
faster. It happens that i switched implementation languages and got a
10-fold performance increase, but i then got a 1,000-fold performance
increase from other techniques.

Having said that, you could maybe consider Go or even Rust for your
neural network.

Liam

-- 
Liam Quin,B https://www.delightfulcomputing.com/
Available for XML/Document/Information Architecture/XSLT/
XSL/XQuery/Web/Text Processing/A11Y training, work & consulting.
Barefoot Web-slave, antique illustrations: B http://www.fromoldbooks.org