/
changelog.txt
192 lines (128 loc) · 5.07 KB
/
changelog.txt
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
v1.4.4 - 2014-09-06
-------------------
* Fixed k-means random initialization error that caused it to drop some
clusters after the first iteration.
v1.4.3 - 2014-01-15
-------------------
* Fixed some k-means parser errors.
* K-means++ deterministic (det++) now works using a pivoting approach.
v1.4.2 - 2013-11-05
-------------------
* The PNorm metric when p = 2.0 (Euclidean distance) is now much faster.
v1.4.1 - 2013-11-05
-------------------
* LeaderFollower uses similarity weighing.
* Changed 'det++' behavior.
v1.4.0 - 2013-10-14
-------------------
* Correlation has been changed to PearsonsCorrelation.
* SpearmansCorrelation added as metric.
* LeaderFollower may use the following metrics: Cosine, PearsonsCorrelation and
SpearmansCorrelation. Its usage has been simplified.
v1.3.5 - 2013-10-11
-------------------
- Added cosine metric.
- The cluster object now stores the mean spatial centroid.
- The deterministic k-means++ initialization allows the user to set the
first centroid.
- Modified Leader-Follower. It now sorts the voxels by peak time and
amplitude before inspecting them. Also, returns the voxel with the
smallest Euclidean distance to the centroid.
v1.3.4 - 2013-09-23
-------------------
- Changed LeaderFollower behavior.
v1.3.3 - 2013-09-06
-------------------
- Fixed the iterator. Again. Once and for all.
- Fixed a bug in the RealMatrix2IJ method that caused it to shift slices.
v1.3.2 - 2013-09-06
-------------------
- A new bug was introduced in the last update -- first voxel was always returned
by the image iterator. Fixed.
v1.3.1 - 2013-09-06
-------------------
- Implements PlugIn interface instead of PlugInFilter.
- Solved bug in iterator (not returning first voxel) that caused PCA / SVD /ICA
to crash on unmasked images.
v1.3.0 - 2013-08-22
-------------------
- Removed all methods and variables related to noisy TAC skipping.
v1.2.6 - 2013-08-21
-------------------
- Fixed a bug in k-means++ initialization distance computation.
v1.2.5 - 2013-08-08
-------------------
- SVD did not work if the soon-to-be-deprecated (hopefully) skip_noisy was not set to true.
- The logic for not returning masked voxels did not work as expected. Fixed.
- ImagePlusHypIterator rewritten (nasty bug showed up when correct pixel masking was implemented).
v1.2.4 - 2013-08-05
-------------------
- ImagePlusHypIterator does not return masked voxels
- Added isMasked(Voxel v) and isMasked(double [] tac) to MathUtils
- K-means initialization algorithms do not choose masked voxels
- Skip noisy voxels is now set to false (might be deprecated in the future).
- LeaderFollower number of clusters is 50 by default (from 1000).
- ImagePlusHyp is now just a wrapper for ImagePlus, but does not extend it.
This avoids duplicating the memory used by the analyzed image.
v1.2.3 - 2013-07-16
-------------------
- Added "About" panel.
- Included deterministic k-means++ initialization.
- Fixed Leader-Follower bug.
- Added RMSD metric.
v1.2.2 - 2013-03-07
-------------------
- PCA can now be computed on the correlation matrix instead of the covariance.
- Additional images from PCA, ICA and SVD use a floating point format.
v1.2.1 - 2013-03-05
-------------------
- Fixed bug: number of components was ignored in ICA.
v1.2.0 - 2013-03-04
-------------------
- ClusteringTechniques can now return a formatted string array that will be saved as additional information.
- Added ICA implementation thanks to the fastICA library by Michael Lambertz.
- Added SVD implementation.
v1.1.9 - 2013-02-27
-------------------
- Added an option for displaying the PCA image in the PCA technique.
- PCA computation now uses SVD on the covariance matrix instead of obtaining
the eigenvectors manually.
v1.1.8 - 2012-10-22
-------------------
- Fixed yet another major bug in LeaderFollower implementation.
- cluster image now has correct voxel size information.
v1.1.7 - 2012-09-28
-------------------
- Fixed major bug in LeaderFollower implementation.
v1.1.6 - 2012-09-27
-------------------
- Added kmeans++ initialization.
- It is now redistributable as a .jar file thanks to Johannes Schindelin.
v1.1.5 - 2012-09-24
-------------------
- Uses calibration data from the image to return calibrated TACs.
- Use SSE to compute when K-Means should stop.
- Minor bug fixes and improvements.
v1.1.4 - 2012-09-10
-------------------
- Fixed a null pointer problem when saving data from an empty cluster.
- Fixed cluster behavior and updated documentation.
v1.1.3 - 2012-09-06
-------------------
- Added PCA technique.
v1.1.2 - 2012-09-06
-------------------
- ImagePlusHyp now implements Iterable<Vector> and is much easier to go
through all the voxels in an image.
v1.0.2 - 2012-08-31
-------------------
- Mahalanobis distance implemented.
v1.0.1 - 2012-08-31
-------------------
- Leader-follower can now discard the smallest cluster (according to
Cluster.score() method) if cluster limit has been reached.
- New javadoc.
v1.0.0 - 2012-08-31
-------------------
- First released version. Implements K-Means and Leader-follower clustering
techniques and two metrics: correlation and p-norm.