WIP: unify Learner1D, Learner2D and LearnerND - #220

Open

jbweston wants to merge 105 commits into

mainfrom

feature/unify-learners

Commits on Oct 6, 2019

WIP: add new learner concept and example notebook
jbweston
committed
1e633cd
WIP: correct implementation of unfinished point removal
jbweston
committed
a5a8f5f
add necessary information to scale data for loss functions that want to
jbweston
committed
d7cb56c
start internal queue attributes with underscore
jbweston
committed
a288b64
recalculate losses when scale increases by a certain factor
jbweston
committed
f2d6125
blackify
jbweston
committed
5d3c2b1
make 1D points use floats rather than length-1 tuples
jbweston
committed
9db175a
correct point insertion logic
jbweston
committed
cdb669d

Commits on Oct 7, 2019

Commits on Oct 8, 2019

make everything work
jbweston
committed
18a88bc
make everything work again
jbweston
committed
e3fd6cc
update Interval class
jbweston
committed
632faf9
add 'remove' method to Domain
jbweston
committed
bd574ab
implement 'ask(tell_pending=False)'
jbweston
committed
3e27066
flakify
jbweston
committed
772a5f4

Commits on Oct 9, 2019

small refactors and comments
jbweston
committed
053ba24
rename bound_points to boundary_points
jbweston
committed
5e36e24
do not evaluate boundary points in __init__
jbweston
committed
fd6bce4
put pending points in separate data structure from data
jbweston
committed
be07de9
replace '_scaled_loss' with 'priority'
jbweston
committed
d07b9c3
rename 'self.loss' to 'self.loss_function'
jbweston
committed
df667df
store priorities as (priority, loss)
jbweston
committed
5377438
docstring update
jbweston
committed
fec432b
correctly compute point priorities
jbweston
committed
61bf9df
refactor comments and fix minor logic bug
jbweston
committed
ce4d9d2
introduce a dict of losses as an optimization
jbweston
committed
ccabce3
clarify the ordering of queried items and priorities
jbweston
committed
1434f69
don't store a tuple as priority, use 'self.losses' from with 'self.loss'
jbweston
committed
cdc6d49
comments
jbweston
committed
166a912
make points a tuple as they should be
jbweston
committed
ba5d207
correctly update pending points and data
jbweston
committed
52fed32
update example notebook
jbweston
committed
5ea9b56

Commits on Oct 11, 2019

Commits on Oct 12, 2019

correct implementation of queue
jbweston
committed
207ac47
black
jbweston
committed
f1fdfee
fix neighbor detection when adding points
jbweston
committed
bc38a7b
abstract out making new subtriangulations
jbweston
committed
b3c394d
black
jbweston
committed
f38e903
add map from subpoints to subdomains to speed up triangulation splitting
jbweston
committed
0cff467
separate Queue and Domain into separate modules
jbweston
committed
5d5402b
add tests for priority queue and domains
jbweston
committed
95c0381
add hypothesis to test requirements
jbweston
committed
d191e3c
rename 'sub_domains' to 'sub_triangulations'
jbweston
committed
b70eea8
flake
jbweston
committed
4305536
add __all__ to domain.py
jbweston
committed
faf68cc
use math module
jbweston
committed
5377e69
docstring formatting
jbweston
committed
4b7f390
increase allowed time for tests to complete
jbweston
committed
9ed92e8
remove Queue.priorities method
jbweston
committed
6e95d68
Queue.items() returns items in priority order
jbweston
committed
5710762
make LossFunction an ABC
jbweston
committed
3b2d9cd
allow the learner to take a ConvexHull as bounds
jbweston
committed
fc348b2
add more comments to the learner
jbweston
committed
953f62d
show that the new learner works for ND output
jbweston
committed
ebcd6f2
typo
jbweston
committed
6c07070
correct sign in simplex boundary equations
jbweston
committed
64a6d0f
correct abstract loss function interface
jbweston
committed
9624f7f
rename need_loss_udate_factor
jbweston
committed
0055d57
correct boundary point choosing logic
jbweston
committed
cc1f83f
make point adding idempotent in learner
jbweston
committed
19ab936
fix bugs in 1D domain
jbweston
committed
a0bea12
fix bug in tell_pending
jbweston
committed
1a84089
catch edge case in loss
jbweston
committed
bf64362
implement data saving
jbweston
committed
f171173
explicitly set inf
jbweston
committed
6111a7c
add new learner to some tests (those that it passes)
jbweston
committed
4ede935
add xfailing tests for new learner
jbweston
committed
6908ec2
correctly set inf
jbweston
committed
484cebf
replace Domain.__contains__ with Domain.contains_subdomain
jbweston
committed
460bd2e
correct implementation of points outside domain
jbweston
committed
929a9fd
implement method 'Domain.encloses' for finding points inside a domain
jbweston
committed
3696379

Commits on Oct 13, 2019

allow NewLearnerND to accept points not in the domain
jbweston
committed
7ff9372
make sure we can ask for more points after adding data outside the
jbweston
committed
3979872
add new learner to datasaving test
jbweston
committed
bb38de3
correct conditional
jbweston
committed
60e0a0c
make 'points_inside' select points from more than 1 subdomain
jbweston
committed
ceaa87d
correct name of domain test
jbweston
committed
a13711a
rename point loss to point priority
jbweston
committed
fbbe44a
do 1 operation per line only
jbweston
committed
d227e1b
correct docstring
jbweston
committed
9f8774a
correct test that all internal points are reassigned when splitting
jbweston
committed
6bdbe88
run all domain tests over random domains, rather than hypercubes
jbweston
committed
32518d1
update priority queue docstring
jbweston
committed
5d3434d
neater ordering
jbweston
committed
6efc250

Commits on Oct 14, 2019

Commits on Oct 16, 2019

fix bug where we receive more points than we asked for
jbweston
committed
52175f0
add learner tests that simulate the runner and randomly asking/telling
jbweston
committed
cdf26a7
create queue on initialization rather than calling 'insert'
jbweston
committed
011c5f3
suppress context when raising when an item was not in a queue
jbweston
committed
51901cc
fix bug in subdomain updating logic when telling data
jbweston
committed
d401cbb
improve comment
jbweston
committed
445c6f1
blackify domain.py
jbweston
committed
d41b355
check that points matrix is well conditioned
jbweston
committed
7430143

Commits on Oct 17, 2019