Reasoning time issue

[Nikolic, Snezana]

Greetings!

Thank you for attending the meeting and for your comments.

As you probably noticed, we have run into the problem with reasoning
involving complex logical definitions pertaining to dietary intakes of
nutrients.

I have tried using different reasoners (Pellet, Racer, HermiT) on two
different machines, but only HermiT 1.3.x worked and only on very limited
number of data instances.

The logical definitions that I talked about during the meeting have all
been independently tested and they classify the target data correctly. In
the test data scenario, when the ontology contains only the logical
definition, for say "low intake of A", all test data instances that fall
in the "low intake of A" range get classified correctly and reasoning is
completed in minutes time.

However, the problem with very long reasoning time arises when the small
set of data instances of type "intake of A" has to be "evaluated" against
2 or 3 logical definitions (e.g.,"low intake of A", "high intake of A" and
"normal intake of A"), which is the case the closest to the real life
scenario. In such cases as well as in the case when the file contains all
test data instances (app 300 instances) the reasoning takes hours.

The following five files include different combinations of data sets and
logical definitions, with data pertaining to average daily valine intake
(which is one or 9 amino acid intakes currently covered by the relevant
literature). Your time allowing, if you could take a look, try them out on
your machine(s) and provide some comments, that would be very appreciated.


1. OwTestVAL-1LogDef-11data.owl
https://code.google.com/p/onstr/source/browse/data/testDataSets/OwTestVAL-1
LogDef-11data.owl
This file contains ONSTR with the logical definition for "low valine
intake data" and 11 data instances for low valine intake. This is the case
for testing the correctness of this logical definition for "low valine
intake" (ONSTR_ Reasoning takes minutes or less, depending on processor(s).


2. OwTestVAL-3LogDef-11data.owl
https://code.google.com/p/onstr/source/browse/data/testDataSets/OwTestVAL-3
LogDef-11data.owl
This is ONSTR with 11 data instances for low valine intake and 3 logical
definitions (for low/high/normal valine intake). Reasoning takes longer
than in 1) but still in minutes range.

3.OwTestVAL-2LogDef-16data.owl
https://code.google.com/p/onstr/source/browse/data/testDataSets/OwTestVAL-2
LogDef-16data.owl
ONSTR with only 16 test data instances and two logical definitions. (11
instances of low valine intake data + 5 for normal valine intake).
Reasoner takes hours to complete the task.

4. OwTestVAL-3LogDef-33data.owl
https://code.google.com/p/onstr/source/browse/data/testDataSets/OwTestVAL-3
LogDef-33data.owl
ONSTR with 33 test data instances that are "handled" by all three logical
definitions pertaining to average daily valine intake (ONSTR_6097970).
Longer reasoning time than 3) and 4).

5. ONSTR_withTESTdata.owl
https://code.google.com/p/onstr/source/browse/data/testDataSets/ONSTR_withT
ESTdata.owl
This file contains ONSTR with all test data instances, roughly 300 test
data instances for all 9 amino acid intakes and total protein intakes.


Note: In all files, ONSTR contains logical definitions for intakes of all
relevant amino acids and total protein, but these are used in reasoning
(in data classification) only in the case 5).
For those who like to read the .owl file:
In all files, some HTML/XML comments for data instances are not 100%
correct. Please don't rely on them.


If you would be so kind to provide some/any feedback in the next 2-3 weeks
that would be greatly appreciated!


P.S.
If you would need more information about the project, you may take a look
at:

1. The ONSTR ICBO2013 presentation is available at:
https://code.google.com/p/onstr/source/browse/docs/Publications/ICBO2013_ON
STR_NikolicEtAl_FINALpres.pptx

2. Oct 21, 2013 Meeting presentation is available at:
https://code.google.com/p/onstr/source/browse/docs/MeetingNotesAndPresentat
ions/Meeting_2013-10-21.pptx

3. Other NBSDC project related info are available at:
https://nbsdc.org


Thank you in advance!

snez



________________________________

This e-mail message (including any attachments) is for the sole use of
the intended recipient(s) and may contain confidential and privileged
information. If the reader of this message is not the intended
recipient, you are hereby notified that any dissemination, distribution
or copying of this message (including any attachments) is strictly
prohibited.

If you have received this message in error, please contact
the sender by reply e-mail message and destroy all copies of the
original message (including attachments).

[nav...@yahoo.com]

Snez;

Thanks for the long message and the links. Who attended the meeting?
Sorry I missed it.

Sham

Sent from my iPhone

> --
> You received this message because you are subscribed to the Google Groups "onstr-discuss" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to onstr-discus...@googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.

[Melanie Courtot]

Hi Snezana,

I was wondering what is the difference between "low average daily dietary valine intake per body mass" and "low average daily dietary valine intake" and why aren't they related in the hierarchy? It seems a bit odd to have a fully duplicated tree.

Cheers,
Melanie

[Monosij Dutta-Roy]

I wanted to ask if it may be possible to abstract the reason for this. Are any, some or all of the following reasons attributable to this.

A1. Contextuality of incoming data not supported (or incomplete) in reasoning metadata.

A2. Referenced reasoning metadata (basis of the reasoning) does not align with contexts of incoming data. This 'mostly' same as above.

B1 Compute time is excessive, but completes, because of combinations to go through.

B2. Compute is incomplete due to mismatch of incoming and referenced reasoning metadata.

Just wanted to see if it could be generalized.

And if the statements do not apply then no prob.

Thanks.

Mono

--
...
Monosij

[Nikolic, Snezana]

Hi Melanie!

Two series of data items "average daly dietary nutrient A intake" and
"average daly dietary nutrient A intake per body mass" represent two
different types of nutrient intake measurements commonly used in nutrition
practice.

1. The "average daly dietary nutrient A intake" refers to the quantity of
particular nutrient ingested in 24 hours. For instance, person X ingested
(ate) 30 grams of protein in 24 hours. For this type of data instances, in
nutrition practice the commonly used unit is gram per (gram/day or g/d).
Depending on the target nutrient, other units are also used (e.g.,
milligram per day (mg/d) or microgram per day (ug/d).
"Average daly dietary nutrient A intake during observed time interval" is
a grand average of all averages intakes of the intakes of given nutrient
during 1, 2, 3 or 7 days, depending on the time period covered by the diet
record obtained from a patient.


2. The "average daly dietary nutrient A intake per body mass" is different
type of datum, because it refers to the daily intake of specific nutrient
with respect to the persons body mass.
For instance,
- person X body mass is 15kg and person X ate 30 grams of protein in a day
(30 g/d)
- average daly dietary protein intake per body mass for person A (for
specific given time interval) would be "average daily intake of protein"
divided by person's A body mass, that is
"average daly dietary protein intake per body mass" would be 30 / 15000 =
0.0002 of grams per kilogram (of body mass) per day. Units used for this
type of data, depending on given nutrient are

gram per kilogram per day (g/kg/d), milligram per kilogram pet day
(mg/kg/d) and microgram per kilogram per day (ug/kg/d)
"Average daly dietary nutrient A intake per body mass during observed time
interval" is a ratio of "average daly dietary nutrient A intake" and
patients body mass.




I'm sure that Dr. Singh would explain all this in better words, but as far
as I understood, the "trick" with using body mass is to figure out whether
the intake of some nutrient is sufficient with respect to body needs and
in that case, the size (mass of the body) does matter. Roughly put, "30
grams per day of nutrient A" may be enough to satisfy child's need for
this specific nutrient, but the same "30 g/d fo nutrient A" may not be
sufficient for an adult whose body mass is several times greater than that
of a child.

The units used for these intake measurements: g/d, g/kg/d and the like
(e.g., dl/d, l/d, floz/d for liquids), are all nutrition domain specific
units, currently not asserted in UO, to the best of my knowledge.


Also, I didn't really get what exactly did you have in mind by saying "why
are they not related in the hierarchy"
These are two different types of average dietary intake measurements under
one parent class "average daily nutrient intake datum". Did you have in
mind some specific relation?


I hope this helps.


Best,
snez

[Alan Ruttenberg]

Just a quick note to others: I replied off list with some findings from a review of the files Snezana sent out. I didn't copy the list as it wasn't clear that all are interested in the level of detail I was addressing or that it was particularly generalizable. However if there is anyone who does want me to I can send you a summary of the findings. I made some concrete suggestions for edits and fixes for Snezana that I thought would help make progress and look forward to her sending out the next revision.

Best

-Alan

[Bill Duncan]

Can you please share your findings with me?

- Bill

[Sivaram Arabandi, MD]

Alan,
I think there are enough folks who would be interested at this level of detail. Could you please share it? I think it will be easier for folks to ignore things not relevant/interested. :)

cheers
Sivaram