machine-learning.one-hot: Better error checking.

extra/machine-learning/one-hot/one-hot-tests.factor

@@ -12,8 +12,17 @@ CONSTANT: test-data {
     { "well" "sick" "tired" }
 }
 
+! encode { hot sunny light neutral slowly well }
 { { 1 0 0 0 0 1 1 0 0 0 1 0 1 1 0 0 } }
-[ { 0 3 0 2 1 0 } test-data one-hot ] unit-test
+[ test-data { 0 3 0 2 1 0 } one-hot ] unit-test
 
 { { 0 1 0 0 1 0 0 1 0 1 0 1 0 0 1 0 } }
-[ { 1 2 1 1 0 1 } test-data one-hot ] unit-test
+[ test-data { 1 2 1 1 0 1 } one-hot ] unit-test
+
+! need an index for each category, e.g. 6 indices
+[ test-data { 1 2 } one-hot ]
+[ one-hot-length-mismatch? ] must-fail-with
+
+! last category is not within { well sick tired }
+[ test-data { 1 2 1 1 0 10 } one-hot ]
+[ one-hot-input-out-of-bounds? ] must-fail-with

extra/machine-learning/one-hot/one-hot.factor

@@ -1,11 +1,25 @@
 ! Copyright (C) 2012 Doug Coleman.
 ! See http://factorcode.org/license.txt for BSD license.
-USING: arrays kernel math.statistics math.vectors sequences
+USING: arrays kernel math math.statistics math.vectors sequences
 sequences.extras ;
 IN: machine-learning.one-hot
 
-: one-hot ( indices features -- array )
+ERROR: one-hot-length-mismatch vcategories vinput ;
+
+ERROR: one-hot-input-out-of-bounds vcategories vinput ;
+
+: check-one-hot-length ( vcateories vinput -- vcategories vinput )
+    2dup [
+        [ length ] bi@ = [ one-hot-length-mismatch ] unless
+    ] [
+        v- [ 1 < ] any?
+        [ one-hot-input-out-of-bounds ] when
+    ] 2bi ;
+
+: one-hot ( features indices -- array )
     [ 1 ] 2dip
-    [ length ] map
-    [ cum-sum0 v+ ]
-    [ nip sum 0 <array> ] 2bi [ set-nths ] keep ;
+    [ [ length ] map ] dip
+    check-one-hot-length
+    [ [ cum-sum0 ] dip v+ ]
+    [ drop sum 0 <array> ] 2bi
+    [ set-nths ] keep ;