-
Notifications
You must be signed in to change notification settings - Fork 10
/
ChangeLog
220 lines (156 loc) · 8.16 KB
/
ChangeLog
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
## NOTE: "->" indicates changes to computation or parameters that affect results
-------------------------------------------------------------------------------------
2014-08-07 Eero Simoncelli <[email protected]>
-> changed params.amplitude.kderange from [0.3 1.1] to [0.3 1.0]
(should be less than 1.0)
* AmplitudeThresholdGUI: added buttons, colors (to identify
cells), spike counts, inf => eliminate cell
2014-08-05 Eero Simoncelli <[email protected]>
* Offset channels in all displays
* spikesort_demo/cbp_spikesort_demo_script.m:
- Fixed bug: CBP step 2, recalculation of waveforms, was operating
on all spikes, not thresholded spikes.
2014-07-30 Eero Simoncelli <[email protected]>
* spikesort_demo/load_raw_data.m:
-> Changed default waveform_len for Harris data set from 81 to 35.
* spikesort_demo/preproc/OrderWaveformsByNorm.m
Modified order of waveforms: was originally sorted from smallest
to largest norm. Now reversed (largest to smallest).
* spikesort_demo/preproc/PartitionSignal.m (PartitionSignal):
Changed back to original thresholding rule:
params.partition.silence_threshold*std(rms)
2014-07-23 Eero Simoncelli <[email protected]>
* spikesort_demo/load_default_parameters.m:
set default filter.freq to [100 5000]
added cbp.lambda, cbp.reweight_exp, cbp.reweight_fn
* spikesort_demo/cbp_spikesort_demo_script.m:
added display of CBP recovered spikes, histograms of residual
errors, PC scatterplot
2014-07-21 Eero Simoncelli <[email protected]>
* spikesort_demo/load_default_parameters.m:
Eliminated params.whitening.p_norm parameter, which was not being used
-> Changed default params.partition.silence_threshold parameter from 3 to 4
* spikesort_demo/load_raw_data.m:
-> Changed default filtering cutoff for Harris data set from 800Hz to 400Hz
2014-07-18 Eero Simoncelli <[email protected]>
* spikesort_demo/load_default_parameters.m:
removed params.cbp.num_features. Unnecessary - can get it from length(init_features)
* spikesort_demo/SpikesortCBP.m:
Added waveforms argument to this function, and removed the cbp_outer.init_features parameter
* spikesort_demo/preproc/plotting/VisualizeClustering.m:
indicate centroids in scatter plot
display of waveforms/traces for multi-electrodes offset
match color scheme in clustering scatter and waveform plots
use same y-axis scaling for all waveforms
* spikesort_demo/preproc/EstimateInitialWaveforms.m (peak_idx):
princomp is being discontinued. Switched to call pca
-> don't re-center (subtract mean) when doing PCA
make waveform_len odd
* spikesort_demo/preproc/FilterData.m (data):
added histograms of raw data
2014-07-17 Peter Li
* put another Harris dataset into the example_data directory
* re-activated greedy solution for speedup
2014-07-16 Eero Simoncelli <[email protected]>
* spikesort_demo/preproc/PartitionSignal.m (signal_rms):
fixed bug introduced yesterday when changing parameterization...
2014-07-15 Eero Simoncelli <[email protected]>
* spikesort_demo/preproc/plotting/VisualizeClustering.m:
got rid of the second fignum argument (always make them consecutive)
added threshold argument
* spikesort_demo/preproc/PartitionSignal.m (snippet_centers):
partition.threshold -> partition.silence_threshold
partition.min_separation_len -> partition.min_silence_len
* spikesort_demo/preproc/EstimateInitialWaveforms.m (threshold):
-> Changed to just use threshold as is (compared directly to rms over channels);
* spikesort_demo/load_default_parameters.m:
-> clustering.threshold -> clustering.spike_threshold, value changed to 6
2014-07-14 Eero Simoncelli <[email protected]>
* spikesort_demo/spikesort_demo_setup.m:
Switched matlabpool calls to use parpool (matlabpool is being discontinued)
* spikesort_demo/cbp_spikesort_demo_script.m (params):
Moved plotting code into functions LoadRawData, FilterData, and PreprocessTrace
2014-07-06 Eero Simoncelli <[email protected]>
* spikesort_demo/load_default_parameters.m (adjust_wfsize_fn):
Moved cbp and cbp_outer parameters into here.
* spikesort_demo/preproc/PartitionSignal.m (dead_zone_idx):
Modified threshold code: threshold parameter is now in units of stdev.
* spikesort_demo/preproc/PreprocessTrace.m (noise_zone_idx):
- commented out display of pre-whitened marginal histograms.
These are not relevant, since we currently hardwire the exponent
to 2 in the CBP optimization.
* spikesort_demo/load_default_parameters.m:
- Moved num_waveforms back into clustering section (hadn't realized
there was a separate one for CBP!)
- Moved default partitioning parameters into here.
2014-07-05 Eero Simoncelli <[email protected]>
* spikesort_demo/load_raw_data.m:
-> Changed default Quiroga filtering.freq back to []
Current results on Quiroga dataset 1, with waveforms re-fit twice,
and threshold of cell 1 slightly hand-adjusted to eliminate
refractory violations:
Spike counts: [1921;1473;1138]
Clustering: Misses=241/3477 (6.93%) FPs=1531/4767 (32.12%)
CBP: Misses=3/3477 (0.09%) FPs=2/3476 (0.06%)
2014-07-03 Eero Simoncelli <[email protected]>
* spikesort_demo/spikesort_demo_setup.m:
error-protected the call to matlabpool, so that it now just gives
a warning.
* spikesort_demo/preproc/PreprocessTrace.m (noise_zone_idx):
noise_threshold parameter moved from general to whitening section.
* spikesort_demo/load_raw_data.m:
- re-organized loading of data files and specialization of
parameters. Now it's all done in this single function, which
replaces load_data_defaults.m
* spikesort_demo/load_default_parameters.m:
added plotting.first_fig_num, and changed all display code to use
figure numbers offset from this number
default fontsize changed to 12
num_waveforms parameter moved from clustering to general section
* spikesort_demo/preproc/plotting/VisualizeClustering.m (nh):
plot noise ball
* spikesort_demo/preproc/EstimateInitialWaveforms.m (threshold):
-> changed threshold for choosing segments for clustering to
mean + 4*std of a chi distribution (about 3.9)
* ToDo and ChangeLog: moved to main directory
* spikesort_demo/preproc/PreprocessTrace.m (nchan):
only plot cross-corr for multi-channel recordings
Display: fig3: histogram of norms, with threshold
fig4: ACF of noise regions
fig5: cross-channel correlations, but only if more than one channel!
fig6: noise distribution, compared with Gaussian and Laplacian
2014-07-02 Eero Simoncelli <[email protected]>
* cbp_spikesort_demo_script.m:
- moved top-level description text into README.md file. Added ecos
citation, and brief outline of algorithm.
- fixed bug: reweight_exp - this used to be hardcoded to three
entries
- further cleanup in plotting of histograms of filtered data
2014-07-01 Eero Simoncelli <[email protected]>
* load_data_defaults.m:
-> added 100Hz cutoff to Quiroga dataset
* cbp_spikesort_demo_script.m (dataMag):
- indicate filtering cutoffs in frequency plot
- plot chi2 distribution for comparison to data magnitude
* preproc/FilterData.m (Wn): allow params.filtering.freq to have a
single entry (low-freq cutoff) for a high-pass filter.
2014-06-19 Eero Simoncelli <[email protected]>
* preproc/PreprocessTrace:
- moved mean subtraction, and global rescaling operations into FilterData.m
- got rid of optional argument parsing for handling plot options, and used params.general.plot_diagnostics instead
2014-06-12 Eero Simoncelli <[email protected]>
* cbp_spikesort_demo_script:
- rewrote comments to describe algorithm steps
- added diagnotistic figures for filtering / whitening
- added plots of Fourier magnitudes
- added comments indicating what to verify in plots, and how to adjust parameters if needed
* load_default_parameters:
- added general.plot_diagnostics
* preproc/FilterData
- pad_size -> pad
- freq_range -> freq
Peter's results on Quiroga data
Spike counts: [1150;1187;1135]
Elapsed time is 71.133004 seconds.
Clust Misses=256/3477 (7.36 percent) FPs=4/3225 (0.12 percent)
CBP Misses=13/3477 (0.37 percent) FPs=2/3466 (0.06 percent)