FrankWolfe.jl

FrankWolfe v0.1.9

Diff since v0.1.8

Closed issues:

• In-place stochastic gradient interface (#233)

Merged pull requests:

• Polynomials example: Change how the true polynomial is generated (#242) (@alejandro-carderera)
• in-place stochastic Frank-Wolfe (#243) (@matbesancon)
• L inf normball (#246) (@j-geuter)
• improve docstring for shifted scaled L1 (#247) (@matbesancon)

FrankWolfe v0.1.8

Diff since v0.1.7

Merged pull requests:

• fix iterator sense (#240) (@matbesancon)

FrankWolfe v0.1.7

Diff since v0.1.6

Closed issues:

• Optimize active set management (#158)
• Stochastic Frank-Wolfe batch size as iterator (#230)

Merged pull requests:

• Literate 1 (#226) (@j-geuter)
• fix activate, update Readme (#228) (@matbesancon)
• updating, fixing formatting readme (#229) (@matbesancon)
• batch and momentum iterators (#231) (@matbesancon)
• make official iterator functions, add to docs (#234) (@matbesancon)
• freeze CSV version (#236) (@matbesancon)
• more iterations on stoch opt (#237) (@matbesancon)
• Bcg example (#239) (@j-geuter)

FrankWolfe v0.1.6

Diff since v0.1.5

Closed issues:

• DataType changes in Blended Conditional Gradients (#181)
• one of the examples in CI blows up (#219)

Merged pull requests:

• Second example (#211) (@j-geuter)
• use similar(x), test generic (#216) (@matbesancon)
• added gamma to trajectory state (#217) (@matbesancon)
• added gamma to lcg (#218) (@pokutta)
• rank-one matrix unary operator (#221) (@matbesancon)
• added scaling for rank-one matrices (#222) (@matbesancon)
• Fixed nasty bug in rankone arithmetic (#223) (@pokutta)
• Third example (#224) (@j-geuter)
• added precision for F64 (#225) (@matbesancon)
• Bound normballs (#227) (@matbesancon)

FrankWolfe v0.1.5

Diff since v0.1.4

Merged pull requests:

• First example (#200) (@j-geuter)
• fix example MOI (#201) (@matbesancon)
• fixed lcg reporting issue (#204) (@pokutta)
• Adding tests for the afw and bcg precision conversion issue (#205) (@pokutta)
• [WIP] adding verbose iterations reporting - DO NOT MERGE (#206) (@pokutta)
• [WIP] Fix for issue of Daniel with adaptive line search (#207) (@pokutta)
• DO NOT MERGE: backup of cc code from Daniel's problem (#208) (@pokutta)
• undoing merge mess (#209) (@pokutta)
• allow more flexible mutable arrays (#215) (@matbesancon)

FrankWolfe v0.1.4

Diff since v0.1.3

Closed issues:

• add extended tests for all examples (#189)

Merged pull requests:

• [WIP] BATS style testing for examples (#196) (@pokutta)
• [WIP] Precompilation of common signatures (#197) (@pokutta)
• [WIP] fixing CI issue for older julia with version dep code (#198) (@pokutta)
• Docs home page and examples.md renamed (#199) (@j-geuter)

FrankWolfe v0.1.3

Diff since v0.1.2

Closed issues:

• Logging (#2)
• [Low] add basic readme (#9)
• [Low] in the end have a note that explains things (#10)
• homogenize logging of output (#140)
• Add time limits to the algorithms (#177)
• fix ugly reporting of step size strategy (#180)
• make sure that x0 is reported as very first iterate. (#187)
• Several examples broken due to trajectory callback (#188)

Merged pull requests:

• Docs branch with basic documentation structure (#184) (@j-geuter)
• Homogenous timeout (#185) (@matbesancon)
• fix LS display (#186) (@matbesancon)
• Reset timing for x_0 (#190) (@pokutta)
• [WIP] Logging callback function (#192) (@pokutta)
• documenter removed from main deps (#194) (@matbesancon)
• formatter, bump version (#195) (@matbesancon)

FrankWolfe v0.1.2

Diff since v0.1.1

Closed issues:

• Registering package (#174)

Merged pull requests:

• monotonous steps (#178) (@matbesancon)
• docstring on algorithms, cleanup NEP (#182) (@matbesancon)
• Time out feature (#183) (@matbesancon)

FrankWolfe v0.1.1

Diff since v0.1.0

First release with arxiv