[MRG] LogisticRegression convert to float64 (sag)#9020

Closed

Henley13 wants to merge 26 commits intoscikit-learn:masterfrom

Henley13:is/8769

Commits on Jun 16, 2017

Remove unused code
Joan Massich
committed
Squash all the PR 9040 commits

Imbert Arthur
authored and
Joan Massich
committed
Append 64 when the templated type is a double
Joan Massich
committed

Commits on Jun 19, 2017

Add templating
Joan Massich
committed
Add missing stuff
Joan Massich
committed
Fix tempita
Joan Massich
committed
Working templated version of sag_fast
Joan Massich
committed
32 64 bit _finite macros (there should be a better way)
Joan Massich
committed
Import sag64 as sag to mimic the default behavior
Joan Massich
committed
add saga and group the asserts
Joan Massich
committed
allow direct coef inspection from the pudb
Joan Massich
committed
test 64 bits first
Joan Massich
committed
Use [float64, float32] as default
Joan Massich
committed
Allow sag to be 64 or 32
Joan Massich
committed
Change the parent function types so that single floats remain 64 like python type
Joan Massich
committed

Commits on Jun 20, 2017

fix np_types
Joan Massich
committed
relax the almost equal assert up to 4 decimals
Joan Massich
committed
test newton-cg and saga
Joan Massich
committed
Relax almost equal only for SAGA case
Joan Massich
committed
make the sag call more readable
Joan Massich
committed
Revert 3e25392
Joan Massich
committed
update gitignore
Joan Massich
committed
Address @raghavrv comments
Joan Massich
committed

Commits on Jun 26, 2017

add a test for y

Imbert Arthur
authored and
Joan Massich
committed

Commits on Jun 29, 2017