home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!cs.utexas.edu!uwm.edu!lll-winken!overload.lbl.gov!dog.ee.lbl.gov!porpoise!manta!usenet
- From: key@blacksea.csc.com (Gerry Key)
- Newsgroups: comp.ai.neural-nets
- Subject: Question: Neural Nets and Data Verification - Reposted
- Keywords: Neural Networks NN Data Verification Environmental
- Message-ID: <1992Aug12.171637.27897@nosc.mil>
- Date: 12 Aug 92 17:16:37 GMT
- Sender: usenet@nosc.mil (Network News)
- Distribution: usa
- Organization: Naval Ocean Systems Center, San Diego, CA
- Lines: 51
-
- Sorry about that last posting ... it got garbled somewhere along the line.
- Here's what it should have said:
-
- I am a biologist and not a regular reader of this newsgroup. Please
- respond directly via email. I will post a summary to comp.ai.neural-nets
- if the responses warrant. I poked around in the FAQs before posting this
- inquiry, but didn't find anything that clearly addressed my question.
-
- I recently saw (but did not read and now cannot find) an article on using
- neural network technology to verify large data matrices. Is anyone
- familiar with this technology and can you give me some background
- information, product names, etc.?
-
- The application I have in mind is environmental survey data - everything
- from fish counts to water chemistry to sediment grain size. Environmental
- databases, or the data matrices derived from them, tend to be large and
- complex, and checking them is tedious work and therefore error-prone.
-
- I'm not so much concerned with numerical outliers that should be
- identified and evaluated during routine statistical analysis. Rather, I'm
- looking for a tool that can check for internal consistency - "quality
- control", if you will. I'd also like a tool that can be used by the
- researcher (rather than requiring a programmer) and one that can be taught
- to recognize similar patterns in different data. I'm not interested in
- conducting neural network research, so am looking for a commercial,
- off-the-shelf (COTS) product with at least nominal vendor support.
-
- A simple-minded example of the application I have in mind would be if the
- variable measured during a survey was salinity, then the units associated
- with that measurement should be parts-per-thousand, g/Kg, or Meq/L, but
- not pH, degrees Celsius, or number per square meter. Similarly, the
- method used to count fish might be a visual transect, beach seine, or a
- fish trap, but not a pH meter, gas chromatograph or sediment sieve.
-
- I recognize that to some extent these tests can be performed by integrity
- checks available through the load module of an RDBMS. However, the
- "rules" why some measurements need closer examination can be quite complex
- and subtle and not easily expressed as if-then-else logic.
-
- The data will come from a variety of sources, some instrumented and some
- not, but all will ultimately wind up in an SQL database. Thus, a tool
- that could be run against such a database would be ideal.
-
- Any suggestions/pointers would be greatly appreciated.
-
- --Gerry Key
-
- Computer Sciences Corporation
- Internet: key@csc.com
- FAX: 619.225.2502
- Voice: 619.225.2504
-