Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Identix (IDNX) -- Ignore unavailable to you. Want to Upgrade?

To: David who wrote (4762)	11/3/1997 7:24:00 PM
From: brad greene	Read Replies (2) \| Respond to of 26039

David, I agree......This stuff gets very technical. From years of listening to Randy, I get the impression that Randy feels that IDX has something special in their patented algorithms. As he explains it, The IDX algorithms enable more to be done and is more efficient than any other possible algorithm .......the perfect equation. He has likened it to finding a needle in a haystack. He has said many times that there is only one best algorithm in his opinion....and that IDX has it. I have always had a great respect for Fowler.....and I very much like what he has done for IDX so far....so I will stand behind his judgement and vision. No big surprise there. It seems to me that IDX must answer the charges in the lawsuit within the next couple of days....Is that the drift you get? Any legal eagle could step in here.......I did not see in the docket any evidence of IDX filing a motion for dismissal. Don't they have to ask for the case to be thrown out?.....or is that just a given...with the judge deciding if and when a case becomes invalid. bg

To: David who wrote (4762)	11/3/1997 8:30:00 PM
From: Jerry Zinser	Read Replies (2) \| Respond to of 26039

Sorry, but there are no barriers (yet) to entry in the commercial markets and in uses of templates or aggregate patterns rather than minutiae. Odds are that every single one of the new finger image biometrics providers is using pattern or template based systems. FBI requires minutiae-based comparisons and matchings for its systems. This is because no (repeat, NO) pattern-based systems comes close to minutiae-based systems in matching. This is true even at the level of "small" data bases for fairly small cities. (Note that such databases are, however, much, much larger than databases for time & attendance, for example.) A statement that it is a privacy advantage to use a pattern or template for a commercial system is a lovely example of trying to make lemonade when you've been given a lemon. In the AFIS systems for law enforcement, scanners (such as produced by IDX and others) are merely an input subsystem. The AFIS provider determines what algorithms will be used for matching (and, usually, for determining what minutiae to use in matches). There are two major barrier to entry into the AFIS matching market. First, each AFIS supplier (NEC, Morpho, NEC, Cogent (with TRW), and now Lockheed-Martin) has proprietary matching algorithms. For competitive reasons, Printrak has been trying to push for open exchange of algorithms. The systems provided by one AFIS supplier are absolutely incompatible in format with those provided by another AFIS supplier. Any scanner meeting FBI specs can, however, be accepted. The second barrier, quite major, is that each AFIS system (city, state, national) has its own database of fingerprint images, processed for efficiency in matching for that particular system. No customer who is even remotely happy with its AFIS supplier wants to face the idea of converting its database. So it is hard, or very expensive, for one AFIS supplier to replace another AFIS supplier at any given customer. The FBI is having LM, as part of their AFIS-FBI contract, be responsible for converting all of the FBI files of fingerprint images into digital files. The FBI receives requests from States. A State sends a new image (live scan or on cards or as captured from latent prints at a crime scene, etc.) and FBI as the center of last resort tries to find a match. Most claims of accuracy or error rates, etc. must be viewed with very healthy skepticism. There may soon be some objective tests and measures, with results available to those who need them, but today the standards are not present. The Law Enforcement arena addresses this problem by requiring benchmark tests and measurements as part of the evaluation of a prospective new system. Because of FBI work, the accuracy and errors of minutiae-based systems are becoming generally known across law enforcement agencies. Pattern and template systems are inherently less precise. They are almost always much more rapid, however, and they almost always require much less storage per finger. Does the lack of precision and accuracy matter? For a lot of things it doesn't matter. Imagine, however, that a credit card company wants to compare the finger image for a defaulting card with the finger images from all other cards to ensure that the deadbeat (or felonist) has not obtained other cards in other names (stolen identities, for example) at other addresses. The inherent errors in a pattern or template system would make it very, very difficult to do such matching (called one-against-many matching). If 100 people are on an authorized list to open a secured door, using pattern matching as part of a 2-level (name plus image) should be more than accurate (as long as you aren't securing a door to Fort Knox, for which the potential value of a crime would make it worthwhile for criminals to find a finger image coming close enough to a match of the images of one of the named persons. So you would have to add more levels of information to check. Hope this helps your understanding of some of the technical forces in your markets.