3 Specimen Classifier Simulator
4 -----------------------------
5 Comments relate to ghc, version 0.19
6 ------------------------------------
11 07-04-94 ipoole added pi method for Lognum
17 To run, type "SpecClass". The program prints a table (take several minutes).
21 1) SpecClass.lhs seems to take ~10 times longer to compile than any
22 of the other modules, even though it is not the biggest. (Without
23 optimisation it seems normal).
25 2) Execution appears to be ~ 4 times slower than interpreted gofer!
26 This is almost certainly due to the functions toDouble and toLognum
27 which in Haskell have to go via Rational. What is clearly needed
28 is a method of type Real a => a -> Double, which could then be
29 defined efficiently for each Real type. The default method
30 could still go via Rational. I don't understand how
31 Haskell has survived for so long without such a method (which
32 makes me fear I've missed something...).
34 3) When compiled *without* optimsation, the program now (7-04-94)
37 4) A very hacky "showFixed" is used, to force Reals < 0.0 to be displayed
38 as 0.XXXXX, rather than in scientific form. There must be better
41 5) The source is not pretty.
43 6) To run with Gofer, you'll need our Haskellised Gofer prelude, which I
44 can provide (it originated from Andy Gill).
46 7) Anyone interested in the statistics underlying the classifier model should
47 request: Ian Poole, "A statistical model for classifying cervical
48 monolayers", RN94_003, MRC Human Genetics Unit, Edinburgh.