Decision tree training with difference subsets of training samples based on a plurality of classifications
View Patent ↗A method may include obtaining a plurality of training samples with a plurality of classifications that include a first classification and a second classification, training an initial tree with an initial set of training samples selected from the plurality of training samples using an initial set of feature values extracted from the set of training samples, and, in response to determining that the initial tree incorrectly classified the initial set of training samples at an output node of the initial tree, training a subsequent tree using a subsequent set of feature values extracted from a subsequent set of training samples.
1. A method comprising:
obtaining a plurality of training samples with a plurality of classifications that include a first classification and a second classification;
training an initial tree with an initial set of training samples selected from the plurality of training samples using an initial set of feature values extracted from the set of training samples, wherein the initial set of training samples correspond to the plurality of classifications; and
in response to determining that the initial tree incorrectly classified the initial set of training samples at an output node of the initial tree,
selecting a subsequent set of training samples from the plurality of training samples, and
training a subsequent tree using a subsequent set of feature values extracted from the subsequent set of training samples, the subsequent tree comprising a root node that replaces the output node of the initial tree that incorrectly classified the initial set of training samples,
wherein the subsequent set of training samples correspond to the plurality of classifications, and
wherein the subsequent set of training samples is different from the initial set of training samples.
2. The method of claim 1 , further comprising:
training the initial tree with a first portion of the plurality training samples corresponding to the first classification and a second portion of the plurality of training samples corresponding to the second classification.
3. The method of claim 2 , further comprising:
training the subsequent tree with a subsequent first portion of the plurality of training samples corresponding to the first classification and a subsequent second portion of the plurality of training samples corresponding to the second classification, wherein the subsequent first portion is disjoint from the subsequent second portion.
4. The method of claim 1 , further comprising:
classifying a sample to generate a sample classification with a machine learning model, the machine learning model comprising a decision tree that comprises the initial tree and the subsequent tree.
5. The method of claim 4 , further comprising:
classifying the sample by extracting the initial set of feature values without extracting the subsequent set of feature values.
6. The method of claim 1 , further comprising:
classifying a sample to generate a sample classification to identify that the sample includes an image of a face of a person.
7. The method of claim 1 , further comprising
training the initial tree, wherein the initial set of feature values correspond to Haar-like features that categorize subsections of an image.
8. A system comprising:
a processor;
a memory; and
the memory comprising an application, wherein the application executes on the processor, uses the memory, and is configured for:
classifying a sample to generate a sample classification with a machine learning model, the machine learning model comprising a decision tree that comprises an initial tree and a subsequent tree, wherein the machine learning model is trained by:
obtaining a plurality of training samples with a plurality of classifications that include a first classification and a second classification,
training an initial tree with an initial set of training samples selected from the plurality of training samples using an initial set of feature values extracted from the set of training samples, wherein the initial set of training samples correspond to the plurality of classifications, and
in response to determining that the initial tree incorrectly classified the initial set of training samples at an output node of the initial tree, selecting a subsequent set of training samples from the plurality of training samples and training a subsequent tree using a subsequent set of feature values extracted from the subsequent set of training samples,
wherein the subsequent tree comprises a root node that replaces the output node of the initial tree that incorrectly classified the initial set of training samples,
wherein the subsequent set of training samples correspond to the plurality of classifications, and
wherein the subsequent set of training samples is different from the initial set of training samples.
9. The system of claim 8 , wherein the initial set of training samples includes a first portion of the plurality of training samples corresponding to the first classification and includes a second portion of the plurality of training samples corresponding to the second classification.
10. The system of claim 9 , wherein the subsequent set of training samples includes a subsequent first portion of the plurality of training samples corresponding to the first classification and includes a subsequent second portion of the plurality of training samples corresponding to the second classification.
11. The system of claim 8 , wherein the first portion is disjoint from the second portion.
12. The system of claim 8 , wherein the application is further configured for:
classifying the sample by extracting the initial set of feature values without extracting the subsequent set of feature values.
13. The system of claim 8 , wherein the application is further configured for:
classifying a sample to generate a sample classification to identify that the sample includes an image of a face of a person.
14. The system of claim 8 , wherein the initial set of feature values correspond to Haar-like features that categorize subsections of an image.
15. A set of one or more non-transitory computer readable mediums comprising computer readable program code for:
obtaining a plurality of training samples with a plurality of classifications that include a first classification and a second classification;
training an initial tree with an initial set of training samples selected from the plurality of training samples using an initial set of feature values extracted from the set of training samples, wherein the initial set of training samples correspond to the plurality of classifications; and
in response to determining that the initial tree incorrectly classified the initial set of training samples at an output node of the initial tree, selecting a subsequent set of training samples from the plurality of training samples and training a subsequent tree using a subsequent set of feature values extracted from the subsequent set of training samples,
wherein the subsequent tree comprises a root node that replaces the output node of the initial tree that incorrectly classified the initial set of training samples, the subsequent set of training samples in the plurality of training samples,
wherein the subsequent set of training samples correspond to the plurality of classifications, and
wherein the subsequent set of training samples is different from the initial set of training samples.
16. The set of one or more non-transitory computer readable mediums of claim 15 , further comprising computer readable program code for:
training the initial tree with a first portion of the plurality of training samples corresponding to the first classification and a second portion of the plurality of training samples corresponding to the second classification.
17. The set of one or more non-transitory computer readable mediums of claim further comprising computer readable program code for:
training the subsequent tree with a subsequent first portion of the plurality of training samples corresponding to the first classification and a subsequent second portion of the plurality of training samples corresponding to the second classification, wherein the subsequent first portion is disjoint from the subsequent second portion.
18. The set of one or more non-transitory computer readable mediums of claim further comprising computer readable program code for:
classifying a sample to generate a sample classification with a machine learning model, the machine learning model comprising a decision tree that comprises the initial tree and the subsequent tree.
19. The set of one or more non-transitory computer readable mediums of claim 18 , further comprising computer readable program code for:
classifying the sample by extracting the initial set of feature values without extracting the subsequent set of feature values.
20. The set of one or more non-transitory computer readable mediums of claim 15 , further comprising computer readable program code for:
classifying a sample to generate a sample classification to identify that the sample includes an image of a face of a person.