You can Log In or Sign Up to try out the site.

Evaluations


Actions: Export Results

Evaluation Batch #33 (PUBLIC)

Benchmark:
Quiroga2004 - Difficult 2
Description:
Mw869H <a href="http://gxatvvjtmaaj.com/">gxatvvjtmaaj</a>, [url=http://nxaocclihrfe.com/]nxaocclihrfe[/url], [link=http://dombfkprmzsr.com/]dombfkprmzsr[/link], http://lrkizxtkyhvq.com/
Algorithm:
PCA/P2P with GMM clustering
Author:
belevtsoff
Date Created:
Sept. 6, 2012

True positive: 62.34%

False positive: 32.85%

Benchmark:
Quiroga2004 - Difficult 2
Trial:
Difficult2_noise005
Task State:
Success
TOGGLE DETAILS

Evaluation Summary:

Detection Errors Classification Errors
False Positives False Negatives
Total Total Non-Overlaps Overlaps Total Non-Overlaps Overlaps
12 174 0 174 1093 948 145

Evaluation Results:

GT Unit Found Unit Ground Truth Found Spikes of Unit False Positives False Negatives
(True Spikes) (True Positives) Other Spikes Noise Found by other Unit Not detected
Total N-O O Total N-O O N-O O FP N-O O N-O O
GT = Ground Truth, O = Overlaps, N-O = Non-Overlaps
00001 2 1120 967 153 1016 967 49 948 56 0 0 56 0 48
00002 3 1109 939 170 80 1 79 0 85 12 938 59 0 32
00003 1 1135 958 177 1001 948 53 0 4 0 10 30 0 94

Evaluation Plots

For every neuron in the sorting a piece of data is cut around every of its spikes. This is done for every channel (for multielectrode data) individually. The plot shows all cut spike waveforms superimposed over each other (gray traces). Dashed lines indicate channel boundaries. Colored waveforms represent the average of all spike waveforms (the template) for each neuron.

All spike waveforms superimposed.

The projections of all spikes onto the first two principle components is shown. Colors indicate neuron identity. This plot gives an impression on how the clusters look like and how good their separation (in PCA space) is. To compute this plot principle component analysis (PCA) is run on all spike waveforms of the sorting. The projections of each waveform is computed on the first two principle components.

This is the same as the previous cluster plot but for PCs 3 and 4.

For each pair of neurons the projections of every spike of both neurons on the vector that connects the templates is shown. This plot is described in Pouzat et al. 2002 "Using noise signature to optimize spike-sorting and to assess neuronal classification quality" (fig. 3 and 6) but here, the noise covariance matrix is not taken into account. Colors indicate neuron identity. The plot gives an impression on how well each pair of clusters is separable. Note however, that the uploaded spike sorting was used to compute this plot, the true separability using the ground truth could be different.

The first second of the spike trains of the sorting are plotted. This plot can be used to see if the website interpreted the uploaded spike train file correctly. Also, if the spike sorter splitted one cluster incorrectly into two (e.g. due to waveform change over time) this is clearly visible in this plot.

True positive: 63.03%

False positive: 32.96%

Benchmark:
Quiroga2004 - Difficult 2
Trial:
Difficult2_noise01
Task State:
Success
TOGGLE DETAILS

Evaluation Summary:

Detection Errors Classification Errors
False Positives False Negatives
Total Total Non-Overlaps Overlaps Total Non-Overlaps Overlaps
2 141 0 141 1139 1027 112

Evaluation Results:

GT Unit Found Unit Ground Truth Found Spikes of Unit False Positives False Negatives
(True Spikes) (True Positives) Other Spikes Noise Found by other Unit Not detected
Total N-O O Total N-O O N-O O FP N-O O N-O O
GT = Ground Truth, O = Overlaps, N-O = Non-Overlaps
00001 3 1187 1060 127 1109 1054 55 1021 44 0 6 44 0 28
00002 2 1136 1007 129 44 3 41 1 59 2 1004 47 0 41
00003 1 1139 998 141 1029 981 48 5 9 0 17 21 0 72

Evaluation Plots

For every neuron in the sorting a piece of data is cut around every of its spikes. This is done for every channel (for multielectrode data) individually. The plot shows all cut spike waveforms superimposed over each other (gray traces). Dashed lines indicate channel boundaries. Colored waveforms represent the average of all spike waveforms (the template) for each neuron.

All spike waveforms superimposed.

The projections of all spikes onto the first two principle components is shown. Colors indicate neuron identity. This plot gives an impression on how the clusters look like and how good their separation (in PCA space) is. To compute this plot principle component analysis (PCA) is run on all spike waveforms of the sorting. The projections of each waveform is computed on the first two principle components.

This is the same as the previous cluster plot but for PCs 3 and 4.

For each pair of neurons the projections of every spike of both neurons on the vector that connects the templates is shown. This plot is described in Pouzat et al. 2002 "Using noise signature to optimize spike-sorting and to assess neuronal classification quality" (fig. 3 and 6) but here, the noise covariance matrix is not taken into account. Colors indicate neuron identity. The plot gives an impression on how well each pair of clusters is separable. Note however, that the uploaded spike sorting was used to compute this plot, the true separability using the ground truth could be different.

The first second of the spike trains of the sorting are plotted. This plot can be used to see if the website interpreted the uploaded spike train file correctly. Also, if the spike sorter splitted one cluster incorrectly into two (e.g. due to waveform change over time) this is clearly visible in this plot.

True positive: 62.21%

False positive: 32.88%

Benchmark:
Quiroga2004 - Difficult 2
Trial:
Difficult2_noise015
Task State:
Success
TOGGLE DETAILS

Evaluation Summary:

Detection Errors Classification Errors
False Positives False Negatives
Total Total Non-Overlaps Overlaps Total Non-Overlaps Overlaps
0 169 3 166 1131 996 135

Evaluation Results:

GT Unit Found Unit Ground Truth Found Spikes of Unit False Positives False Negatives
(True Spikes) (True Positives) Other Spikes Noise Found by other Unit Not detected
Total N-O O Total N-O O N-O O FP N-O O N-O O
GT = Ground Truth, O = Overlaps, N-O = Non-Overlaps
00001 3 1142 966 176 56 7 49 14 44 0 959 82 0 45
00002 1 1113 952 161 1007 931 76 927 70 0 20 39 1 46
00003 2 1185 1029 156 1077 1010 67 55 21 0 17 14 2 75

Evaluation Plots

For every neuron in the sorting a piece of data is cut around every of its spikes. This is done for every channel (for multielectrode data) individually. The plot shows all cut spike waveforms superimposed over each other (gray traces). Dashed lines indicate channel boundaries. Colored waveforms represent the average of all spike waveforms (the template) for each neuron.

All spike waveforms superimposed.

The projections of all spikes onto the first two principle components is shown. Colors indicate neuron identity. This plot gives an impression on how the clusters look like and how good their separation (in PCA space) is. To compute this plot principle component analysis (PCA) is run on all spike waveforms of the sorting. The projections of each waveform is computed on the first two principle components.

This is the same as the previous cluster plot but for PCs 3 and 4.

For each pair of neurons the projections of every spike of both neurons on the vector that connects the templates is shown. This plot is described in Pouzat et al. 2002 "Using noise signature to optimize spike-sorting and to assess neuronal classification quality" (fig. 3 and 6) but here, the noise covariance matrix is not taken into account. Colors indicate neuron identity. The plot gives an impression on how well each pair of clusters is separable. Note however, that the uploaded spike sorting was used to compute this plot, the true separability using the ground truth could be different.

The first second of the spike trains of the sorting are plotted. This plot can be used to see if the website interpreted the uploaded spike train file correctly. Also, if the spike sorter splitted one cluster incorrectly into two (e.g. due to waveform change over time) this is clearly visible in this plot.

True positive: 60.38%

False positive: 34.76%

Benchmark:
Quiroga2004 - Difficult 2
Trial:
Difficult2_noise02
Task State:
Success
TOGGLE DETAILS

Evaluation Summary:

Detection Errors Classification Errors
False Positives False Negatives
Total Total Non-Overlaps Overlaps Total Non-Overlaps Overlaps
21 191 18 173 1193 1053 140

Evaluation Results:

GT Unit Found Unit Ground Truth Found Spikes of Unit False Positives False Negatives
(True Spikes) (True Positives) Other Spikes Noise Found by other Unit Not detected
Total N-O O Total N-O O N-O O FP N-O O N-O O
GT = Ground Truth, O = Overlaps, N-O = Non-Overlaps
00001 3 1151 989 162 87 67 20 89 43 1 916 84 6 58
00002 1 1195 1035 160 992 924 68 799 65 15 108 48 3 44
00003 2 1147 989 158 1030 951 79 165 32 5 29 8 9 71

Evaluation Plots

For every neuron in the sorting a piece of data is cut around every of its spikes. This is done for every channel (for multielectrode data) individually. The plot shows all cut spike waveforms superimposed over each other (gray traces). Dashed lines indicate channel boundaries. Colored waveforms represent the average of all spike waveforms (the template) for each neuron.

All spike waveforms superimposed.

The projections of all spikes onto the first two principle components is shown. Colors indicate neuron identity. This plot gives an impression on how the clusters look like and how good their separation (in PCA space) is. To compute this plot principle component analysis (PCA) is run on all spike waveforms of the sorting. The projections of each waveform is computed on the first two principle components.

This is the same as the previous cluster plot but for PCs 3 and 4.

For each pair of neurons the projections of every spike of both neurons on the vector that connects the templates is shown. This plot is described in Pouzat et al. 2002 "Using noise signature to optimize spike-sorting and to assess neuronal classification quality" (fig. 3 and 6) but here, the noise covariance matrix is not taken into account. Colors indicate neuron identity. The plot gives an impression on how well each pair of clusters is separable. Note however, that the uploaded spike sorting was used to compute this plot, the true separability using the ground truth could be different.

The first second of the spike trains of the sorting are plotted. This plot can be used to see if the website interpreted the uploaded spike train file correctly. Also, if the spike sorter splitted one cluster incorrectly into two (e.g. due to waveform change over time) this is clearly visible in this plot.