User Reference:P300Classifier - Revision history

Mellinger: /* Classifier Weight Normalization */

2026-06-23T13:34:56Z

Classifier Weight Normalization

← Older revision		Revision as of 13:34, 23 June 2026
Line 255:		Line 255:

	==Classifier Weight Normalization==		==Classifier Weight Normalization==
	When applying classifier weights to obtain a classification score, and ~~select~~ a target with the largest score, this procedure is invariant to multiplication of all classifier weights with a constant factor. We are, thus, free to normalize classifier weights with a factor that allows us to interpret classification scores in a certain way.		When applying classifier weights to obtain a classification score, and when selecting a target with the largest score, this procedure is invariant against multiplication of all classifier weights with a constant factor.
			We are, thus, free to normalize classifier weights with a factor that allows us to interpret classification scores in a certain way.

	As described [[Programming_Reference:StimulusTask_Class#Accumulation_and_Combination_of_Evidence\|here]], a normalization that makes within-class variance identical to unity ~~when the classifier is applied to the training data set~~ will result in classification scores to represent an estimate of the log evidence in favor of the respective target. This allows the AccumulateEvidence and MinimumEvidence parameters to work.		As described [[Programming_Reference:StimulusTask_Class#Accumulation_and_Combination_of_Evidence\|here]], a normalization that makes within-class variance identical to unity will result in classification scores to represent an estimate of the log evidence in favor of the respective target. This allows the [[Programming_Reference:StimulusTask_Class#Accumulation_and_Combination_of_Evidence\|AccumulateEvidence and MinimumEvidence parameters]] to work.

	==Future work==		==Future work==

Mellinger: /* Classifier Weight Normalization */

2026-06-23T12:39:23Z

Classifier Weight Normalization

← Older revision		Revision as of 12:39, 23 June 2026
Line 256:		Line 256:
	==Classifier Weight Normalization==		==Classifier Weight Normalization==
	When applying classifier weights to obtain a classification score, and select a target with the largest score, this procedure is invariant to multiplication of all classifier weights with a constant factor. We are, thus, free to normalize classifier weights with a factor that allows us to interpret classification scores in a certain way.		When applying classifier weights to obtain a classification score, and select a target with the largest score, this procedure is invariant to multiplication of all classifier weights with a constant factor. We are, thus, free to normalize classifier weights with a factor that allows us to interpret classification scores in a certain way.

	As described [[Programming_Reference:StimulusTask_Class#Accumulation_and_Combination_of_Evidence\|here]], a normalization that makes within-class variance identical to unity when the classifier is applied to the training data set will result in classification scores to represent an estimate of the log evidence in favor of the respective target. This allows the AccumulateEvidence and MinimumEvidence parameters to work.		As described [[Programming_Reference:StimulusTask_Class#Accumulation_and_Combination_of_Evidence\|here]], a normalization that makes within-class variance identical to unity when the classifier is applied to the training data set will result in classification scores to represent an estimate of the log evidence in favor of the respective target. This allows the AccumulateEvidence and MinimumEvidence parameters to work.

Mellinger: /* Future work */

2026-06-23T12:38:44Z

Future work

← Older revision		Revision as of 12:38, 23 June 2026
Line 260:		Line 260:
	==Future work==		==Future work==

	* Present classification results as topographies and graphs as the Matlab-based P300 GUI (has been done in BCI2000Analysis)		* Present classification results as topographies and graphs as the Matlab-based P300 GUI (has been done in [[User_Reference:BCI2000Analysis\|BCI2000Analysis]])

	* Let the P300Classifier GUI work with frequency-based features (has been done in BCI2000Analysis)		* Let the P300Classifier GUI work with frequency-based features (has been done in [[User_Reference:BCI2000Analysis\|BCI2000Analysis]])

	* Speed up the P300Classifier GUI by using multithreading when solving the Least Squares (QR factorization)		* Speed up the P300Classifier GUI by using multithreading when solving the Least Squares (QR factorization)

Mellinger: /* Batch Scripting */

2025-10-28T12:46:45Z

Batch Scripting

← Older revision		Revision as of 12:46, 28 October 2025
Line 107:		Line 107:
	-inicfg C:\BCI2000.x64\tools\P300Classifier\parameters.ini		-inicfg C:\BCI2000.x64\tools\P300Classifier\parameters.ini
	-ClassifierOutputFile C:\BCI2000.x64\data\ClassifierTest001\P3Classifier_ChS1_RAW_SW.prm		-ClassifierOutputFile C:\BCI2000.x64\data\ClassifierTest001\P3Classifier_ChS1_RAW_SW.prm
	-~~PscoreDumpTraining~~ C:\BCI2000.x64\data\ClassifierTest001\PscoreTraining.txt		-PscoreDumpFileTraining C:\BCI2000.x64\data\ClassifierTest001\PscoreTraining.txt
	-~~PscoreDumpTesting~~ C:\BCI2000.x64\data\ClassifierTest001\PscoreTesting.txt		-PscoreDumpFileTesting C:\BCI2000.x64\data\ClassifierTest001\PscoreTesting.txt
	</pre>		</pre>

Mellinger: /* Batch Scripting */

2025-10-28T12:44:28Z

Batch Scripting

← Older revision		Revision as of 12:44, 28 October 2025
Line 100:		Line 100:

	===Batch Scripting===		===Batch Scripting===
	The P300 Classifier Tool can be scripted in a batch file using the example format below:		The P300 Classifier Tool can be scripted in a batch file using the example format below (note this needs to be entered as a single line):

	<pre>		<pre>

Mellinger: /* Batch Scripting */

2025-10-28T12:43:54Z

Batch Scripting

← Older revision		Revision as of 12:43, 28 October 2025
Line 103:		Line 103:

	<pre>		<pre>
	C:\BCI2000.x64\tools\P300Classifier\P300Classifier.exe -TrainingDataFiles C:\BCI2000.x64\data\ClassifierTest001\ClassifierTestS001R01.dat		C:\BCI2000.x64\tools\P300Classifier\P300Classifier.exe
			-TrainingDataFiles C:\BCI2000.x64\data\ClassifierTest001\ClassifierTestS001R01.dat
	-inicfg C:\BCI2000.x64\tools\P300Classifier\parameters.ini		-inicfg C:\BCI2000.x64\tools\P300Classifier\parameters.ini
	-ClassifierOutputFile C:\BCI2000.x64\data\ClassifierTest001\P3Classifier_ChS1_RAW_SW.prm		-ClassifierOutputFile C:\BCI2000.x64\data\ClassifierTest001\P3Classifier_ChS1_RAW_SW.prm

Mellinger: /* Batch Scripting */

2025-10-28T12:43:40Z

Batch Scripting

← Older revision		Revision as of 12:43, 28 October 2025
Line 103:		Line 103:

	<pre>		<pre>
	C:\BCI2000.x64\tools\P300Classifier\P300Classifier.exe -TrainingDataFiles C:\BCI2000.x64\data\ClassifierTest001\ClassifierTestS001R01.dat -inicfg C:\BCI2000.x64\tools\P300Classifier\parameters.ini -ClassifierOutputFile C:\BCI2000.x64\data\ClassifierTest001\P3Classifier_ChS1_RAW_SW.prm		C:\BCI2000.x64\tools\P300Classifier\P300Classifier.exe -TrainingDataFiles C:\BCI2000.x64\data\ClassifierTest001\ClassifierTestS001R01.dat
			-inicfg C:\BCI2000.x64\tools\P300Classifier\parameters.ini
			-ClassifierOutputFile C:\BCI2000.x64\data\ClassifierTest001\P3Classifier_ChS1_RAW_SW.prm
	-PscoreDumpTraining C:\BCI2000.x64\data\ClassifierTest001\PscoreTraining.txt		-PscoreDumpTraining C:\BCI2000.x64\data\ClassifierTest001\PscoreTraining.txt
	-PscoreDumpTesting C:\BCI2000.x64\data\ClassifierTest001\PscoreTesting.txt		-PscoreDumpTesting C:\BCI2000.x64\data\ClassifierTest001\PscoreTesting.txt

Mellinger: /* Batch Scripting */

2025-10-28T12:43:16Z

Batch Scripting

← Older revision		Revision as of 12:43, 28 October 2025
Line 104:		Line 104:
	<pre>		<pre>
	C:\BCI2000.x64\tools\P300Classifier\P300Classifier.exe -TrainingDataFiles C:\BCI2000.x64\data\ClassifierTest001\ClassifierTestS001R01.dat -inicfg C:\BCI2000.x64\tools\P300Classifier\parameters.ini -ClassifierOutputFile C:\BCI2000.x64\data\ClassifierTest001\P3Classifier_ChS1_RAW_SW.prm		C:\BCI2000.x64\tools\P300Classifier\P300Classifier.exe -TrainingDataFiles C:\BCI2000.x64\data\ClassifierTest001\ClassifierTestS001R01.dat -inicfg C:\BCI2000.x64\tools\P300Classifier\parameters.ini -ClassifierOutputFile C:\BCI2000.x64\data\ClassifierTest001\P3Classifier_ChS1_RAW_SW.prm
			-PscoreDumpTraining C:\BCI2000.x64\data\ClassifierTest001\PscoreTraining.txt
			-PscoreDumpTesting C:\BCI2000.x64\data\ClassifierTest001\PscoreTesting.txt
	</pre>		</pre>

Mellinger: /* Classifier Weight Normalization */

2025-05-06T20:44:55Z

Classifier Weight Normalization

← Older revision		Revision as of 20:44, 6 May 2025
Line 251:		Line 251:
	==Classifier Weight Normalization==		==Classifier Weight Normalization==
	When applying classifier weights to obtain a classification score, and select a target with the largest score, this procedure is invariant to multiplication of all classifier weights with a constant factor. We are, thus, free to normalize classifier weights with a factor that allows us to interpret classification scores in a certain way.		When applying classifier weights to obtain a classification score, and select a target with the largest score, this procedure is invariant to multiplication of all classifier weights with a constant factor. We are, thus, free to normalize classifier weights with a factor that allows us to interpret classification scores in a certain way.
	As described [[Programming_Reference:StimulusTask_Class#Accumulation_and_Combination_of_Evidence\|here]], a normalization that makes within-class variance identical to unity when the classifier is applied to the training data set will result in classification scores to represent an estimate the log evidence in favor of the respective target. This allows the AccumulateEvidence and MinimumEvidence parameters to work.		As described [[Programming_Reference:StimulusTask_Class#Accumulation_and_Combination_of_Evidence\|here]], a normalization that makes within-class variance identical to unity when the classifier is applied to the training data set will result in classification scores to represent an estimate of the log evidence in favor of the respective target. This allows the AccumulateEvidence and MinimumEvidence parameters to work.

	==Future work==		==Future work==

Mellinger: /* Classifier Weight Normalization */

2025-05-06T20:44:31Z

Classifier Weight Normalization

← Older revision		Revision as of 20:44, 6 May 2025
Line 251:		Line 251:
	==Classifier Weight Normalization==		==Classifier Weight Normalization==
	When applying classifier weights to obtain a classification score, and select a target with the largest score, this procedure is invariant to multiplication of all classifier weights with a constant factor. We are, thus, free to normalize classifier weights with a factor that allows us to interpret classification scores in a certain way.		When applying classifier weights to obtain a classification score, and select a target with the largest score, this procedure is invariant to multiplication of all classifier weights with a constant factor. We are, thus, free to normalize classifier weights with a factor that allows us to interpret classification scores in a certain way.
	As described [Programming_Reference:StimulusTask_Class#Accumulation_and_Combination_of_Evidence here], a normalization that makes within-class variance identical to unity when the classifier is applied to the training data set will result in classification scores to represent an estimate the log evidence in favor of the respective target. This allows the AccumulateEvidence and MinimumEvidence parameters to work.		As described [[Programming_Reference:StimulusTask_Class#Accumulation_and_Combination_of_Evidence\|here]], a normalization that makes within-class variance identical to unity when the classifier is applied to the training data set will result in classification scores to represent an estimate the log evidence in favor of the respective target. This allows the AccumulateEvidence and MinimumEvidence parameters to work.

	==Future work==		==Future work==