Functional Connectivity (Neural Network Method) - Revision history

Chris at 19:22, 12 June 2019

2019-06-12T19:22:25Z

← Older revision		Revision as of 15:22, 12 June 2019
Line 1:		Line 1:
	Some classes of neural networks are well-suited for analyzing functional connectivity because they are sensitive to patterns of co-occurrences within the input data. If the data are composed of fMRI activations across a brain regions, then weights within the network that encodes these activations will encode the regularity with which regions coactivate. In [http://online.liebertpub.com/doi/pdf/10.1089/brain.2013.0174 McNorgan & Joanisse (2014)], we demonstrated that a neural network can arrive at a similar functional connectivity solution to the conventional cross-correlational approach.		Some classes of neural networks are well-suited for analyzing functional connectivity because they are sensitive to patterns of co-occurrences within the input data. If the data are composed of fMRI activations across a brain regions, then weights within the network that encodes these activations will encode the regularity with which regions coactivate. In [http://online.liebertpub.com/doi/pdf/10.1089/brain.2013.0174 McNorgan & Joanisse (2014)], we demonstrated that an autoencoder neural network can arrive at a similar functional connectivity solution to the conventional cross-correlational approach.

	== Preparing Data ==		== Preparing Data ==

Chris: /* Preparing Data */

2016-08-05T19:13:20Z

Preparing Data

← Older revision		Revision as of 15:13, 5 August 2016
Line 4:		Line 4:
	Node activations in a neural network fall within a range of 0 (off) to 1 (on). They can take on intermediate values (and mathematically, 0 and 1 are the theoretical limits that the activation values never actually reach), but generally speaking, values close to these limits are easy to work with and understand; anything in between is a complication. Consequently, you are going to need to transform your fMRI data into values between 0 and 1. A function, <code>binarizeMatrix</code> can be found in the ubfs Scripts/Matlab/ folder. This function takes as a parameter the normalized matrix or cell array, '''''Z'''''. '''''Z''''' will likely be the product of running <code>normalizeMatrix()</code> on one or more time series that were obtained by a call to <code>loadFSTS()</code>. The function <code>binarizeMatrix</code> takes one mandatory parameter, '''''Z''''', and two optional parameters, ''thresh'', and ''clip'':		Node activations in a neural network fall within a range of 0 (off) to 1 (on). They can take on intermediate values (and mathematically, 0 and 1 are the theoretical limits that the activation values never actually reach), but generally speaking, values close to these limits are easy to work with and understand; anything in between is a complication. Consequently, you are going to need to transform your fMRI data into values between 0 and 1. A function, <code>binarizeMatrix</code> can be found in the ubfs Scripts/Matlab/ folder. This function takes as a parameter the normalized matrix or cell array, '''''Z'''''. '''''Z''''' will likely be the product of running <code>normalizeMatrix()</code> on one or more time series that were obtained by a call to <code>loadFSTS()</code>. The function <code>binarizeMatrix</code> takes one mandatory parameter, '''''Z''''', and two optional parameters, ''thresh'', and ''clip'':
	%I have generally found fMRI activations to be typically normally-distributed		%I have generally found fMRI activations to be typically normally-distributed
	thresh=1.~~286~~; %corresponds to a Z-score at the 90th percentile		thresh=1.28; %corresponds to a Z-score at the 90th percentile
	clip=3; %values above/below ±3 are outliers to be clipped (set to NaN)		clip=3; %values above/below ±3 are outliers to be clipped (set to NaN)
	[BIN, SCALED]=binarizeMatrix(Z, thresh, clip);		[BIN, SCALED]=binarizeMatrix(Z, thresh, clip);

	If not set, ''thresh'' defaults to 1.0, and ''clip'' defaults to ∞ (i.e. no values are clipped). Other reasonable values for ''thresh'' include 0 (or, technically, <code>realmin</code>, so we don't have a div/0 error), which binarizes the values into above vs. below average, or critical Z score values~~, such as 1.286, 1.645, 1.96, etc~~. The output matrix, '''''BIN''''', is a binarized transformation of '''''Z''''', where all values above/below ±''thresh'' are set to 1 and 0, respectively, and other values are set to ''NaN''. The matrix '''''SCALED''''' is a non-binary transformation where the intermediate values are set to decimal values between 1 and 0, and where the mean of the time series for each region is 0.5. When training networks using supervised learning, target activations must be either 1 or 0, and so the '''''BIN''''' matrix is appropriate for training the network on target region activations. Input values, however, may take on intermediate values, and so the '''''SCALED''''' matrix might be appropriate to generate input patterns.		If not set, ''thresh'' defaults to 1.0, and ''clip'' defaults to ∞ (i.e. no values are clipped). Other reasonable values for ''thresh'' include 0 (or, technically, <code>realmin</code>, so we don't have a div/0 error), which binarizes the values into above vs. below average, or critical Z score values. The output matrix, '''''BIN''''', is a binarized transformation of '''''Z''''', where all values above/below ±''thresh'' are set to 1 and 0, respectively, and other values are set to ''NaN''. The matrix '''''SCALED''''' is a non-binary transformation where the intermediate values are set to decimal values between 1 and 0, and where the mean of the time series for each region is 0.5. When training networks using supervised learning, target activations must be either 1 or 0, and so the '''''BIN''''' matrix is appropriate for training the network on target region activations. Input values, however, may take on intermediate values, and so the '''''SCALED''''' matrix might be appropriate to generate input patterns.

			Critical Z-score values:
			*0.842 (p=0.2)
			*1.282 (p=0.1)
			*1.644 (p=0.05)
			*2.054 (p=0.02)
			*2.326 (p=0.01)

	== Other Stuff ==		== Other Stuff ==

192.168.1.1: /* Other Stuff */

2016-07-14T16:58:45Z

Other Stuff

← Older revision		Revision as of 12:58, 14 July 2016
Line 15:		Line 15:
	[[Category: Neural Networks]]		[[Category: Neural Networks]]
	[[Category: Network Analyses]]		[[Category: Network Analyses]]
			[[Category: Functional Connectivity]]

192.168.1.1: /* Other Stuff */

2016-07-13T19:32:57Z

Other Stuff

← Older revision		Revision as of 15:32, 13 July 2016
Line 14:		Line 14:

	[[Category: Neural Networks]]		[[Category: Neural Networks]]
			[[Category: Network Analyses]]

192.168.1.1 at 19:30, 13 July 2016

2016-07-13T19:30:37Z

← Older revision		Revision as of 15:30, 13 July 2016
Line 9:		Line 9:

	If not set, ''thresh'' defaults to 1.0, and ''clip'' defaults to ∞ (i.e. no values are clipped). Other reasonable values for ''thresh'' include 0 (or, technically, <code>realmin</code>, so we don't have a div/0 error), which binarizes the values into above vs. below average, or critical Z score values, such as 1.286, 1.645, 1.96, etc. The output matrix, '''''BIN''''', is a binarized transformation of '''''Z''''', where all values above/below ±''thresh'' are set to 1 and 0, respectively, and other values are set to ''NaN''. The matrix '''''SCALED''''' is a non-binary transformation where the intermediate values are set to decimal values between 1 and 0, and where the mean of the time series for each region is 0.5. When training networks using supervised learning, target activations must be either 1 or 0, and so the '''''BIN''''' matrix is appropriate for training the network on target region activations. Input values, however, may take on intermediate values, and so the '''''SCALED''''' matrix might be appropriate to generate input patterns.		If not set, ''thresh'' defaults to 1.0, and ''clip'' defaults to ∞ (i.e. no values are clipped). Other reasonable values for ''thresh'' include 0 (or, technically, <code>realmin</code>, so we don't have a div/0 error), which binarizes the values into above vs. below average, or critical Z score values, such as 1.286, 1.645, 1.96, etc. The output matrix, '''''BIN''''', is a binarized transformation of '''''Z''''', where all values above/below ±''thresh'' are set to 1 and 0, respectively, and other values are set to ''NaN''. The matrix '''''SCALED''''' is a non-binary transformation where the intermediate values are set to decimal values between 1 and 0, and where the mean of the time series for each region is 0.5. When training networks using supervised learning, target activations must be either 1 or 0, and so the '''''BIN''''' matrix is appropriate for training the network on target region activations. Input values, however, may take on intermediate values, and so the '''''SCALED''''' matrix might be appropriate to generate input patterns.

			== Other Stuff ==
			Shazam!

			[[Category: Neural Networks]]

192.168.1.1: /* Preparing Input Data */

2016-07-12T18:34:49Z

Preparing Input Data

← Older revision		Revision as of 14:34, 12 July 2016
Line 1:		Line 1:
	Some classes of neural networks are well-suited for analyzing functional connectivity because they are sensitive to patterns of co-occurrences within the input data. If the data are composed of fMRI activations across a brain regions, then weights within the network that encodes these activations will encode the regularity with which regions coactivate. In [http://online.liebertpub.com/doi/pdf/10.1089/brain.2013.0174 McNorgan & Joanisse (2014)], we demonstrated that a neural network can arrive at a similar functional connectivity solution to the conventional cross-correlational approach.		Some classes of neural networks are well-suited for analyzing functional connectivity because they are sensitive to patterns of co-occurrences within the input data. If the data are composed of fMRI activations across a brain regions, then weights within the network that encodes these activations will encode the regularity with which regions coactivate. In [http://online.liebertpub.com/doi/pdf/10.1089/brain.2013.0174 McNorgan & Joanisse (2014)], we demonstrated that a neural network can arrive at a similar functional connectivity solution to the conventional cross-correlational approach.

	== Preparing ~~Input~~ Data ==		== Preparing Data ==
	~~The input data to the neural network is the time course of fMRI activations from one or more experimental runs.~~ Node activations in a neural network fall within a range of 0 (off) to 1 (on). They can take on intermediate values (and mathematically, 0 and 1 are the theoretical limits that the activation values never actually reach), but generally speaking, values close to these limits are easy to work with and understand; anything in between is a complication. Consequently, you are going to need to transform your fMRI data into values between 0 and 1. A function, <code>binarizeMatrix</code> can be found in the ubfs Scripts/Matlab/ folder. This function takes as a parameter the normalized matrix or cell array, '''''Z'''''. '''''Z''''' will likely be the product of running <code>normalizeMatrix()</code> on one or more time series that were obtained by a call to <code>loadFSTS()</code>. The function <code>binarizeMatrix</code> takes one mandatory parameter, '''''Z''''', and two optional parameters, ''thresh'', and ''clip'':		Node activations in a neural network fall within a range of 0 (off) to 1 (on). They can take on intermediate values (and mathematically, 0 and 1 are the theoretical limits that the activation values never actually reach), but generally speaking, values close to these limits are easy to work with and understand; anything in between is a complication. Consequently, you are going to need to transform your fMRI data into values between 0 and 1. A function, <code>binarizeMatrix</code> can be found in the ubfs Scripts/Matlab/ folder. This function takes as a parameter the normalized matrix or cell array, '''''Z'''''. '''''Z''''' will likely be the product of running <code>normalizeMatrix()</code> on one or more time series that were obtained by a call to <code>loadFSTS()</code>. The function <code>binarizeMatrix</code> takes one mandatory parameter, '''''Z''''', and two optional parameters, ''thresh'', and ''clip'':
	%I have generally found fMRI activations to be typically normally-distributed		%I have generally found fMRI activations to be typically normally-distributed
	thresh=1.286; %corresponds to a Z-score at the 90th percentile		thresh=1.286; %corresponds to a Z-score at the 90th percentile

192.168.1.1: /* Preparing Data */

2016-07-12T18:33:57Z

Preparing Data

← Older revision		Revision as of 14:33, 12 July 2016
Line 1:		Line 1:
	Some classes of neural networks are well-suited for analyzing functional connectivity because they are sensitive to patterns of co-occurrences within the input data. If the data are composed of fMRI activations across a brain regions, then weights within the network that encodes these activations will encode the regularity with which regions coactivate. In [http://online.liebertpub.com/doi/pdf/10.1089/brain.2013.0174 McNorgan & Joanisse (2014)], we demonstrated that a neural network can arrive at a similar functional connectivity solution to the conventional cross-correlational approach.		Some classes of neural networks are well-suited for analyzing functional connectivity because they are sensitive to patterns of co-occurrences within the input data. If the data are composed of fMRI activations across a brain regions, then weights within the network that encodes these activations will encode the regularity with which regions coactivate. In [http://online.liebertpub.com/doi/pdf/10.1089/brain.2013.0174 McNorgan & Joanisse (2014)], we demonstrated that a neural network can arrive at a similar functional connectivity solution to the conventional cross-correlational approach.

	== Preparing Data ==		== Preparing Input Data ==
	Node activations in a neural network fall within a range of 0 (off) to 1 (on). They can take on intermediate values (and mathematically, 0 and 1 are the theoretical limits that the activation values never actually reach), but generally speaking, values close to these limits are easy to work with and understand; anything in between is a complication. Consequently, you are going to need to transform your fMRI data into values between 0 and 1. A function, <code>binarizeMatrix</code> can be found in the ubfs Scripts/Matlab/ folder. This function takes as a parameter the normalized matrix or cell array, '''''Z'''''. '''''Z''''' will likely be the product of running <code>normalizeMatrix()</code> on one or more time series that were obtained by a call to <code>loadFSTS()</code>. The function <code>binarizeMatrix</code> takes one mandatory parameter, '''''Z''''', and two optional parameters, ''thresh'', and ''clip'':		The input data to the neural network is the time course of fMRI activations from one or more experimental runs. Node activations in a neural network fall within a range of 0 (off) to 1 (on). They can take on intermediate values (and mathematically, 0 and 1 are the theoretical limits that the activation values never actually reach), but generally speaking, values close to these limits are easy to work with and understand; anything in between is a complication. Consequently, you are going to need to transform your fMRI data into values between 0 and 1. A function, <code>binarizeMatrix</code> can be found in the ubfs Scripts/Matlab/ folder. This function takes as a parameter the normalized matrix or cell array, '''''Z'''''. '''''Z''''' will likely be the product of running <code>normalizeMatrix()</code> on one or more time series that were obtained by a call to <code>loadFSTS()</code>. The function <code>binarizeMatrix</code> takes one mandatory parameter, '''''Z''''', and two optional parameters, ''thresh'', and ''clip'':
	%I have generally found fMRI activations to be typically normally-distributed		%I have generally found fMRI activations to be typically normally-distributed
	thresh=1.286; %corresponds to a Z-score at the 90th percentile		thresh=1.286; %corresponds to a Z-score at the 90th percentile

192.168.1.1 at 14:24, 7 July 2016

2016-07-07T14:24:03Z

← Older revision		Revision as of 10:24, 7 July 2016
Line 1:		Line 1:
	Some classes of neural networks are well-suited for analyzing functional connectivity because they are sensitive to patterns of co-occurrences within the input data. If the data are composed of fMRI activations across a brain regions, then weights within the network that encodes these activations will encode the regularity with which regions coactivate. In [http://online.liebertpub.com/doi/pdf/10.1089/brain.2013.0174 McNorgan & Joanisse (2014)], I demonstrated that a neural network can arrive at a similar functional connectivity solution to the conventional cross-correlational approach.		Some classes of neural networks are well-suited for analyzing functional connectivity because they are sensitive to patterns of co-occurrences within the input data. If the data are composed of fMRI activations across a brain regions, then weights within the network that encodes these activations will encode the regularity with which regions coactivate. In [http://online.liebertpub.com/doi/pdf/10.1089/brain.2013.0174 McNorgan & Joanisse (2014)], we demonstrated that a neural network can arrive at a similar functional connectivity solution to the conventional cross-correlational approach.

	== Preparing Data ==		== Preparing Data ==

172.101.108.133: /* Preparing Data */

2016-06-30T01:45:11Z

Preparing Data

← Older revision		Revision as of 21:45, 29 June 2016
Line 8:		Line 8:
	[BIN, SCALED]=binarizeMatrix(Z, thresh, clip);		[BIN, SCALED]=binarizeMatrix(Z, thresh, clip);

	If not set, ''thresh'' defaults to 1.0, and ''clip'' defaults to ∞ (i.e. no values are clipped). Other reasonable values for ''thresh'' include 0 (or, technically, <code>realmin</code>, so we don't have a div/0 error), which binarizes the values into above vs. below average, or critical Z score values, such as 1.286, 1.645, 1.96, etc. The output matrix, '''''BIN''''', is a binarized transformation of '''''Z''''', where all values above/below ±''thresh'' are set to 1 and 0, respectively, and other values are set to ''NaN''. The matrix '''''SCALED''''' is a non-binary transformation where the intermediate values are set to decimal values between 1 and 0, and where the mean of the time series for each region is 0.5. When training networks, target activations must be either 1 or 0, and so the '''''BIN''''' matrix is appropriate for training the network on target region activations ~~when using supervised learning~~. Input values may take on intermediate values, and so the '''''SCALED''''' matrix might be appropriate to generate input patterns.		If not set, ''thresh'' defaults to 1.0, and ''clip'' defaults to ∞ (i.e. no values are clipped). Other reasonable values for ''thresh'' include 0 (or, technically, <code>realmin</code>, so we don't have a div/0 error), which binarizes the values into above vs. below average, or critical Z score values, such as 1.286, 1.645, 1.96, etc. The output matrix, '''''BIN''''', is a binarized transformation of '''''Z''''', where all values above/below ±''thresh'' are set to 1 and 0, respectively, and other values are set to ''NaN''. The matrix '''''SCALED''''' is a non-binary transformation where the intermediate values are set to decimal values between 1 and 0, and where the mean of the time series for each region is 0.5. When training networks using supervised learning, target activations must be either 1 or 0, and so the '''''BIN''''' matrix is appropriate for training the network on target region activations. Input values, however, may take on intermediate values, and so the '''''SCALED''''' matrix might be appropriate to generate input patterns.

172.101.108.133: /* Preparing Data */

2016-06-30T01:43:59Z

Preparing Data

← Older revision		Revision as of 21:43, 29 June 2016
Line 8:		Line 8:
	[BIN, SCALED]=binarizeMatrix(Z, thresh, clip);		[BIN, SCALED]=binarizeMatrix(Z, thresh, clip);

	The output matrix, '''''BIN''''', is a binarized transformation of '''''Z''''', where all values above/below ±''thresh'' are set to 1 and 0, respectively, and other values are set to ''NaN''. The matrix '''''SCALED''''' is a non-binary transformation where the intermediate values are set to decimal values between 1 and 0, and where the mean of the time series for each region is 0.5. When training networks, target activations must be either 1 or 0, and so the '''''BIN''''' matrix is appropriate for training the network on target region activations when using supervised learning. Input values may take on intermediate values, and so the '''''SCALED''''' matrix might be appropriate to generate input patterns.		If not set, ''thresh'' defaults to 1.0, and ''clip'' defaults to ∞ (i.e. no values are clipped). Other reasonable values for ''thresh'' include 0 (or, technically, <code>realmin</code>, so we don't have a div/0 error), which binarizes the values into above vs. below average, or critical Z score values, such as 1.286, 1.645, 1.96, etc. The output matrix, '''''BIN''''', is a binarized transformation of '''''Z''''', where all values above/below ±''thresh'' are set to 1 and 0, respectively, and other values are set to ''NaN''. The matrix '''''SCALED''''' is a non-binary transformation where the intermediate values are set to decimal values between 1 and 0, and where the mean of the time series for each region is 0.5. When training networks, target activations must be either 1 or 0, and so the '''''BIN''''' matrix is appropriate for training the network on target region activations when using supervised learning. Input values may take on intermediate values, and so the '''''SCALED''''' matrix might be appropriate to generate input patterns.