[Previous Months][Date Index][Thread Index][Join - Register][Login]   Help@Insulin-Pumpers.org
  [Message Prev][Message Next][Thread Prev][Thread Next]   for subscribe/unsubscribe assistance
 
 

RE: [IP] The new Paradigm Pump w/ continuous glucose monitoring



 No offense, but from this comment, I could probably make a fair guess as to
what TYPE of electrical
 engineer you are, or rather what type you're not. (That's NOT a comment on how
GOOD an engineer you
are! We all end up being specialists.)

 It makes perfect sense to me. I don't know specifically the technology they're
using to manage the
 data, but the pattern of training & calibration makes it look a lot like a
neural network. Neural
 networks (also known as Bayesian networks) must go through a training period,
where a number of
 internal parameters (equivalent to Baysian conditional probabilities) are
incrementally adjusted
until they minimize the error metric (usually RMS error).

 It's not the sensors which are at fault -- it simply requires the collection of
a few days of data
 to arrive at the correct interpretation of the data. I bet the
multi-compartmental nature of the
 problem makes this more complex than simply adjusting a single calibration
parameter.

 No doubt there's considerable variation in individual physiology that requires
such an adaptive
approach.

-----Original Message-----
 From: email @ redacted
[mailto:email @ redacted] On Behalf Of
Jim Skelton
Sent: Tuesday, May 02, 2006 22:27

"In the manual MM wrote that you get the best results with the fourth sensor."

 As an electrical engineer, I must say this sounds like a lame excuse. If the
first three sensors are
crap, then only send out "fourth sensors." .
.
----------------------------------------------------------
for HELP or to subscribe/unsubscribe/change list versions,
contact: HELP@insulin-pumpers.org