Releases: BGU-CS-VIL/DPMMSubClusters.jl
Releases · BGU-CS-VIL/DPMMSubClusters.jl
v0.1.13
DPMMSubClusters v0.1.13
predict now returns '(labels,probability_matrix)' where the matrix is of NxK.
Added smart splits for Gaussians (added keyword to fit 'smart_splits=true').
Bugs:
Closed issues:
- Release 0.1.12 (#24)
Merged pull requests:
v0.1.12
DPMMSubClusters v0.1.12
- Added predict method for an existing model.
Bugs:
Merged pull requests:
- Fix a typo (#22) (@nclarkjudd)
v0.1.11
v0.1.10
DPMMSubClusters v0.1.10
Bugs:
Workaround for numerical errors during NIW posterior calculation.
v0.1.9
DPMMSubClusters v0.1.9
Bugs:
Updated dependencies versions.
Removed redundant dependencies.
Removed redundant code.
Closed issues:
Merged pull requests:
- Install TagBot as a GitHub Action (#15) (@JuliaTagBot)
- support specialfunctions 1.0 (#17) (@stevengj)
v0.1.7
Bugs:
Update functions to SpecialFunctions
0.8
v0.1.6
Bugs:
- Fixed numerical error occurring on high-trials multinomials, which caused the deletion of components
Features:
- Now you can use an outlier component for catching all the outliers, specify its estimated weight, and its hyper params (depends on data and prior, a NIW example will have very high pseudo counts and very wide isotopic variance).
- fit(...) now return the sub labels as well.
- cluster_statistics(points, labels, clusters) is a new function which, given the output of the model (or of other points/labels), provides the avg LL and probability of each cluster/