ahlenius@motcid.UUCP (Mark Ahlenius) (07/07/90)
I am relatively new to nets, and have a basic question on use of nets in the industry. Has there been any work in the area of verification and validation of neural nets - when programmed on conventional computers? I know that this is a tough area for expert systems as well. I realize that the answer to this question this may seem obvious because the nets are trained with examples shown to them, and therefore that is what they are comparing against (sort of). In real world applications, we must carefully verify the design and validate all outcomes - based on potential inputs. Is there any work being done in this area? Are the testing phases just accomplished by showing good and bad examples to a net, and examining the outcomes? Or are there other methods available to "prove" the functioning of the net? With conventional software, much emphasis is going into structured design and testing (even clean room testing) to ensure that companies ship quality products that were designed and built correctly. So I am wondering what will be done (and how) with nets when they start to be delivered as products to customers. thanks in advance 'mark -- =============== regards 'mark ============================================= Mark Ahlenius voice:(708)-632-5346 email: uunet!motcid!ahleniusm Motorola Inc. fax: (708)-632-2413 Arlington, Hts. IL, USA 60004