Difference between revisions of "Talk:Databased advisory System"

From AardRock Wiki
Jump to navigation Jump to search
(Added Martien's feedback.)
 
 
Line 13: Line 13:
* Regarding your objections against Aida:
* Regarding your objections against Aida:
** I suggest both Durk and you discuss this with Harold van der Valk as well.
** I suggest both Durk and you discuss this with Harold van der Valk as well.
= Durk's feedback =
* ''The system forms relations by the input taken for the past.''
** Okay, so your system 'looks at the past'. As this statement is rather general, please elaborate and give some technical details.
* ''The system must be able to make combinations of entries inserted in the past.''
** How does it look at combinations? If a meal is slightly different from another meal, how does the system know if you will still be O.K.?
* ''Het entries of today are compared from those of the past. The result shows to take a different course of action than that of the past. On ... day were your glucose level not in the safety area of 4-8 mmol/l.''
** So the system would be able to say that it went wrong in the past? I seriously doubt if that is very meaningful? Wouldnt it be better to let the system PREVENT trouble/illness instead of waiting untill it goes wrong? Wouldnt it be good to just have zero illness from the moment you use cheetah, instead of going through a year of trouble before the system gets useful?

Latest revision as of 10:33, 15 May 2006

Martien's feedback

  • Nice piece of work Just. Thanks.
  • I like the graph with the increasing uncertainty.
  • I fully agree with the user's familiarity of using a logbook. Cheetah must honour this.
  • I like your suggestion to use user familiar units (like "plate" or "big apple") for users, and absolute values in Cheetah's core and the periodic recalibration. Must have in Cheetah.
  • Seems like Durk, Giel and you must spend some time figuring out how to do this (re-)calibration.
  • You say: "The results have to been shown in a userfriendly format. In that way the user can decide if the recommendations are usefull or that the system made an error. Maybe it is posible to validate the recommendations the system generates by the user.":
    • Cheetah will know it made an error when a future BG level is out of bounds.
    • That BG level is caused by following Cheetah's advice in the first place.
    • This is the most important feedback loop that allows Cheetah to evolve and adapt.
  • You say: "Collecting multiple glucose measurements of persons who ate the same thing, produces a cloud of points."
    • Seems like we need to get in touch with a few scientific people (staticians and medical researchers), probably through Harold van der Valk.
  • Regarding your objections against Aida:
    • I suggest both Durk and you discuss this with Harold van der Valk as well.

Durk's feedback

  • The system forms relations by the input taken for the past.
    • Okay, so your system 'looks at the past'. As this statement is rather general, please elaborate and give some technical details.
  • The system must be able to make combinations of entries inserted in the past.
    • How does it look at combinations? If a meal is slightly different from another meal, how does the system know if you will still be O.K.?
  • Het entries of today are compared from those of the past. The result shows to take a different course of action than that of the past. On ... day were your glucose level not in the safety area of 4-8 mmol/l.
    • So the system would be able to say that it went wrong in the past? I seriously doubt if that is very meaningful? Wouldnt it be better to let the system PREVENT trouble/illness instead of waiting untill it goes wrong? Wouldnt it be good to just have zero illness from the moment you use cheetah, instead of going through a year of trouble before the system gets useful?