[PATCH] tuning.sh

[David Relson]

Greetings,

The patch below incorporates fixes for the problems found by pi and 
Greg.  It also reformats the Top 10 Results as they have suggested and no 
longer tests robs=1 or min_dev=0.475 as these values are just too darn big!

David

--- 01362/tuning/tuning.sh	2003-06-05 13:47:35.000000000 -0400
+++ cvs/tuning/tuning.sh	2003-06-14 15:40:40.000000000 -0400
@@ -11,13 +11,14 @@

  # range of values for testing

-svals="1 3.2e-1 1e-1 3.2e-2 1e-2"
-mvals=`seq 0.025 0.025 0.47501`		# wide value range
+svals="3.2e-1 1e-1 3.2e-2 1e-2"
+mvals=`seq 0.025 0.025 0.45001`		# wide value range
  #mvals=`seq 0.420 0.020 0.46001` 	# high value range

  # approx 0.1%-0.3% of nonspam corpus size for selecting spam_cutoff value

-target=24
+target=12
+export target

  # file names

@@ -70,7 +71,7 @@

  function getco () {
      opts="-m$1 -o$2"
-    shift ; shift ; shift
+    shift ; shift
      res=`cat $* | bogofilter -t -c $CFG $opts -v 2>&1 | \
      perl -e ' $target = $ENV{"target"}; while (<>) { ' \
  	 -e ' ($i, $d) = split; push @diffs, $d unless $i != 1; }' \
@@ -83,7 +84,7 @@
  function wrapper () {
      v="-v"
      opts="-m$1 -o$2"
-    shift ; shift ; shift
+    shift ; shift
      res=`cat $1 | bogofilter -t -c $CFG $opts -v | grep -c $v '^1'`
  }

@@ -93,16 +94,16 @@
      date=`date "+%m/%d %H:%M:%S"`
      echo -n $date "  "
      printf "%-7s %5.3f fpos..." $rs $md
-    getco $md 0.10 $rs r0.ns.mc r1.ns.mc r2.ns.mc
+    getco $md,$rs 0.10 r0.ns.mc r1.ns.mc r2.ns.mc
      fpos=${res##* }; co=${res%% *}; let fpos=$fpos/3
      printf "%d at cutoff %8.6f, run0..." $fpos $co
-    run=0; wrapper $md $co $rs r0.sp.mc; fneg1=$res
+    run=0; wrapper $md,$rs $co r0.sp.mc; fneg1=$res
      echo "$rs $md $co $run $fpos $fneg1" >> $PARM_TBL
      printf "%3d  run1..." $fneg1
-    run=1; wrapper $md $co $rs r1.sp.mc; fneg2=$res
+    run=1; wrapper $md,$rs $co r1.sp.mc; fneg2=$res
      echo "$rs $md $co $run $fpos $fneg2" >> $PARM_TBL
      printf "%3d  run2..." $fneg2
-    run=2; wrapper $md $co $rs r2.sp.mc; fneg3=$res
+    run=2; wrapper $md,$rs $co r2.sp.mc; fneg3=$res
      echo "$rs $md $co $run $fpos $fneg3" >> $PARM_TBL
      printf "%3d"  $fneg3
      let fneg="$fneg1+$fneg2+$fneg3"
@@ -124,7 +125,10 @@

  # get 10 best results (lowest false negative count)
  (echo "" ; \
-echo "Top 10 results" ; \
-grep fpos < $RESULTS | sed "s@\.\.\. at .. @g" | sort -g --key=15 | head -10 
) | tee -a $RESULTS
+echo "Top 10 results:" ; \
+echo " robs   min_dev spam_cutoff  run0 run1 run2 total" ; \
+grep fpos < $RESULTS | grep -v cutoff.0.415000 | sed "s@\.\.\. at .. @g" | \
+awk '{printf "%6.4f %7s %12s  %4d %4d %4d %5d\n", $3, $4, $9, $11, $13, 
$15, $16 }' | \
+tr -d "," | sort -g --key=7 | head -10 ) | tee -a $RESULTS

  date "+%m/%d %H:%M:%S"