/* ---- Google Analytics Code Below */

Saturday, March 28, 2015

On the Architecture and Why of Neural Nets

A good overview of the concepts and key characteristics of neural nets by Stuart Reid.   We used these methods in the enterprise to replace some stat methods, especially when we wanted more control of the method involved.

They were also used where we wanted direct control of nonlinear and complex methods that we knew would address some kinds of data. They worked when we needed to rapidly re-train data in time.

The article is more on the subtleties of why you might use these, rather than the how.    He also mentions some books that address application methods in specific business domains where they are can be useful.   You can build some of these yourself, we did some of that, but its better to use well tested prefabricated systems, there are many out there in R and Python and elsewhere.

No comments: