Online machine learning - Revision history

Suhridk: /* Approaches */

2012-03-24T21:12:45Z

Approaches

2012-03-24T21:11:07Z

Approaches

2012-03-24T21:10:18Z

Stream mining

2012-03-24T20:52:30Z

Stream mining

2012-03-24T20:47:52Z

Stream mining

2012-03-24T20:46:41Z

Stream mining

2012-03-24T20:40:45Z

Stream mining

2012-03-24T20:39:24Z

Stream mining

2012-03-24T20:31:54Z

Stream mining

2012-03-24T20:29:28Z

Stream mining

@@ Line 15: / Line 15: @@
 * Wrapper approach : collect a stream into a batch and then use batch ML. Not very good.
 * Adaptation approach: custom built for stream learning.
-** Decision Tree: Seems to be the best approach.
+** Decision Tree: Good choice.

@@ Line 13: / Line 13: @@
 == Approaches ==
-* Wrapper approach : collect a stream into a batch and then use batch ML.
+* Wrapper approach : collect a stream into a batch and then use batch ML. Not very good.
 * Adaptation approach: custom built for stream learning.

← Older revision		Revision as of 21:10, 24 March 2012
Line 10:		Line 10:
	* The number of examples (rows) are very large and they are assumed to keep arriving in a stream infinitely.		* The number of examples (rows) are very large and they are assumed to keep arriving in a stream infinitely.
	* The labels that can be applied to the data are limited, typically less than ten.		* The labels that can be applied to the data are limited, typically less than ten.
		+
		+	== Approaches ==
		+
		+	* Wrapper approach : collect a stream into a batch and then use batch ML.
		+	* Adaptation approach: custom built for stream learning.

@@ Line 1: / Line 1: @@
 = Stream mining =
-* Core assumption is that '''training''' examples can be briefly inspected for a single time only.
+* Conventional machine learning assumes that the training data is available as a whole set.
 * Arrive in a high speed stream and then must be discarded to make room for subsequent examples.
 * Algorithm must update its model incrementally as each element is inspected.

@@ Line 7: / Line 7: @@
 * the goal of classification is to produce a model that can predict the class of unlabelled examples by training on examples whose label/class is supplied.
 * Data is assumed to have small and fixed no. of columns or attributes/features. This can be thought of as tuples.
 * The labels that can be applied to the data are limited, typically less than ten.

← Older revision		Revision as of 20:40, 24 March 2012
Line 6:		Line 6:
	* It should ideally be able to apply the model any time between training examples.		* It should ideally be able to apply the model any time between training examples.
	* the goal of classification is to produce a model that can predict the class of unlabeled examples by training on examples whose label/class is supplied.		* the goal of classification is to produce a model that can predict the class of unlabeled examples by training on examples whose label/class is supplied.
		+	* Data is assumed to have small and fixed no. of columns or attributes/features. This can be thought of as tuples.

← Older revision		Revision as of 20:39, 24 March 2012
Line 5:		Line 5:
	* Algorithm must update its model incrementally as each element is inspected.		* Algorithm must update its model incrementally as each element is inspected.
	* It should ideally be able to apply the model any time between training examples.		* It should ideally be able to apply the model any time between training examples.
		+	* the goal of classification is to produce a model that can predict the class of unlabeled examples by training on examples whose label/class is supplied.

← Older revision		Revision as of 20:31, 24 March 2012
Line 4:		Line 4:
	* Arrive in a high speed stream and then must be discarded to make room for subsequent examples.		* Arrive in a high speed stream and then must be discarded to make room for subsequent examples.
	* Algorithm must update its model incrementally as each element is inspected.		* Algorithm must update its model incrementally as each element is inspected.
		+	* It should ideally be able to apply the model any time between training examples.

← Older revision		Revision as of 20:29, 24 March 2012
Line 3:		Line 3:
	* Core assumption is that '''training''' examples can be briefly inspected for a single time only.		* Core assumption is that '''training''' examples can be briefly inspected for a single time only.
	* Arrive in a high speed stream and then must be discarded to make room for subsequent examples.		* Arrive in a high speed stream and then must be discarded to make room for subsequent examples.
		+	* Algorithm must update its model incrementally as each element is inspected.