v1.0.3

NMF v1.0.3

Diff since v1.0.2

Merged pull requests:

• New stop condition: use component relative change (#83) (@youdongguo)
• equal comparison between NMF.Results (#84) (@youdongguo)
• Improve inferability (#85) (@timholy)
• Fix verbose printing (#86) (@timholy)

v1.0.2

NMF v1.0.2

Diff since v1.0.1

Merged pull requests:

• SnoopPrecomple -> PrecompileTools (#82) (@timholy)

v1.0.1

NMF v1.0.1

Diff since v1.0.0

Merged pull requests:

• StatsBase0.34 (#81) (@youdongguo)

v1.0.0

NMF v1.0.0

Diff since v0.5.3

Closed issues:

• CoordinateDescent performance vs. sklearn (#43)
• Probabilistic NMF (#61)
• Test fragility (#73)

Merged pull requests:

• Delete REQUIRE (#66) (@ghost)
• Update README.md (#67) (@KwatMDPhD)
• Fix failing test (#69) (@timholy)
• nndsvd: allow initialization with externally-supplied factorization (#70) (@timholy)
• Add precompilation (#71) (@timholy)
• Switch to testsets (#72) (@timholy)
• Use fixed matrices in tests (#74) (@timholy)

v0.5.3

NMF v0.5.3

Diff since v0.5.2

Closed issues:

• Breaking change between 0.5.1 and 0.5.2 (#62)

Merged pull requests:

• Fix the return type of rsvd (#63) (@tsano430)
• Set package version to 0.5.3 (#64) (@tsano430)

v0.5.2

NMF v0.5.2

Diff since v0.5.1

Closed issues:

• Initialization using randomized SVD (#58)

Merged pull requests:

• Remove the code included in NonNegLeastSquares.jl (#57) (@tsano430)
• Replace SVD with randomized SVD (#59) (@tsano430)
• Set package version to 0.5.2 (#60) (@tsano430)

v0.5.1

NMF v0.5.1

Diff since v0.5.0

Closed issues:

• randinit with overly specific arguments (#3)
• Separable NMF algorithm (#11)
• Adjust interface to allow specification of W and H (#19)
• "α is not finite" on CI (#29)
• different scaling factor in nndsvd comparing to original paper (#32)
• Random initialization with normalization not working properly (#34)
• Options for fixing W or H during solve! (#44)
• Incorrect CI status in README.md (#51)

Merged pull requests:

• Improve greedy coordinate descent algorithm (#45) (@tsano430)
• Migrate CI to GitHub Actions (#46) (@ararslan)
• Fix #29: Reimplement alternate least square using projected gradient descent (#47) (@tsano430)
• Fix #19: Support replicates (#48) (@tsano430)
• Fix replicates (#49) (@tsano430)
• Add successive projection algorithm (#50) (@tsano430)
• Fix CI status badges (#52) (@tsano430)
• Add an option for specifying whether to update H (#53) (@tsano430)
• Improve initialization methods (#54) (@tsano430)
• Improve coordinate descent algorithm (#55) (@tsano430)
• Set package version to 0.5.1 (#56) (@tsano430)

v0.5.0

NMF v0.5.0

Diff since v0.4.2

v0.4.2

NMF v0.4.2

Diff since v0.4.1

Closed issues:

• Convergence history (#7)
• :multdiv always gives a NaN matrix and a zero matrix (#15)
• :multmse fails; returns zeros all the time (#16)
• weird MethodError (#28)
• New version tagging (#41)

Merged pull requests:

• fix #16: initialize H in nnmf (#25) (@korenmiklos)
• [Breaking] Add greedy coordinate descent algorithm (#38) (@tsano430)
• Fix#15: Avoidance of zero division (#39) (@tsano430)
• Fix #28: Check the elements of X (#40) (@tsano430)

v0.4.1

NMF v0.4.1

Diff since v0.4.0

Closed issues:

• ETA for Julia 0.7-1.0 compatible release? (#27)
• add init and alg to NMF.Result (#30)
• New version tagging (#35)

Merged pull requests:

• Added the coordinate descent method from scikit-learn, supporting sparsity (#23) (@vilim)
• Updates for 1.0 (#24) (@ararslan)
• Comments for CoordinateDescent method (#33) (@VPetukhov)
• Set package version to 0.4.1 (#37) (@ararslan)