Fixed some grammar and added a note about the bias term.

This commit is contained in:
Davis King 2015-01-05 17:39:37 -05:00
parent a7e55c79e8
commit afa4fe0c23
1 changed files with 11 additions and 10 deletions

View File

@ -18,14 +18,15 @@
best way to measure this goodness isn't obvious and therefore machine learning best way to measure this goodness isn't obvious and therefore machine learning
methods are used. methods are used.
The remainder of this example program will show you how to learn a goodness The remainder of this example will show you how to learn a goodness function
function which is optimal, in a certain sense, for use with the Hungarian which is optimal, in a certain sense, for use with the Hungarian algorithm. To
algorithm. To do this, we will make a simple dataset of example associations do this, we will make a simple dataset of example associations and use them to
and use them to train a supervised machine learning method. train a supervised machine learning method.
Finally, note that there is a whole example program dedicated to assignment learning Finally, note that there is a whole example program dedicated to assignment
problems where you are trying to make an object tracker. So if that is what you are learning problems where you are trying to make an object tracker. So if that is
interested in then read the learning_to_track_ex.cpp example program. what you are interested in then take a look at the learning_to_track_ex.cpp
example program.
*/ */
@ -96,9 +97,9 @@ struct feature_extractor
Recall that our task is to learn the "goodness of assignment" function for Recall that our task is to learn the "goodness of assignment" function for
use with the Hungarian algorithm. The dlib tools assume this function use with the Hungarian algorithm. The dlib tools assume this function
can be written as: can be written as:
match_score(l,r) == dot(w, PSI(l,r)) match_score(l,r) == dot(w, PSI(l,r)) + bias
where l is an element of LHS, r is an element of RHS, w is a parameter vector, where l is an element of LHS, r is an element of RHS, w is a parameter vector,
and PSI() is a user supplied feature extractor. bias is a scalar value, and PSI() is a user supplied feature extractor.
This feature_extractor is where we implement PSI(). How you implement this This feature_extractor is where we implement PSI(). How you implement this
is highly problem dependent. is highly problem dependent.
@ -132,7 +133,7 @@ struct feature_extractor
is "good"). is "good").
!*/ !*/
{ {
// Lets just use the squared difference between each vector as our features. // Let's just use the squared difference between each vector as our features.
// However, it should be emphasized that how to compute the features here is very // However, it should be emphasized that how to compute the features here is very
// problem dependent. // problem dependent.
feats = squared(left - right); feats = squared(left - right);