Skip to content
This repository was archived by the owner on May 21, 2022. It is now read-only.

fix broken tests and fix v0.6 deprecations #3

Merged
Evizero merged 5 commits into from
May 5, 2017
Merged

fix broken tests and fix v0.6 deprecations #3

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
b4458b4
fix broken tests and fix v0.6 deprecations
rdeits May 5, 2017
d37c747
update travis
rdeits May 5, 2017
ed598dc
try to fix requirements
rdeits May 5, 2017
950797d
can't require transformations
rdeits May 5, 2017
d82c515
can't use PenaltyFunctions on 0.5
rdeits May 5, 2017
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion .travis.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,9 +2,9 @@
language: julia
os:
- linux
- osx
julia:
- 0.5
- 0.6
- nightly
notifications:
email: false
Expand Down
1 change: 1 addition & 0 deletions REQUIRE
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,3 +2,4 @@ julia 0.5-

Reexport
LearnBase
Compat 0.17 # for new type syntax
13 changes: 7 additions & 6 deletions src/ObjectiveFunctions.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,6 +2,7 @@ __precompile__(true)

module ObjectiveFunctions

using Compat: @compat
using Reexport
@reexport using LearnBase
@reexport using LossFunctions
Expand All @@ -19,7 +20,7 @@ export
RegularizedObjective,
objective

abstract AbstractLossTransform{T} <: Transformation
@compat abstract type AbstractLossTransform{T} <: Transformation end

immutable NoLoss <: AbstractLossTransform{Void} end

Expand All @@ -41,20 +42,20 @@ immutable LossTransform{T,L<:Loss} <: AbstractLossTransform{T}
target::SumNode{T,1}
output::OutputNode{T,1}

function LossTransform(loss::Loss, nin::Int)
function (::Type{LossTransform{T, L}}){T, L <: Loss}(loss::Loss, nin::Int)
input = InputNode(T, nin)
target = InputNode(T, nin)
output = OutputNode(T, 1)
grad(output)[1] = one(T) # ∂L/∂L == 1
new(loss, nin, input, target, output)
new{T, L}(loss, nin, input, target, output)
end
end

LossTransform{L<:Loss}(loss::L, nin::Int) = LossTransform{Float64, L}(loss, nin)

# input and target are pre-populated... compute the output value as: ∑ loss(targetᵢ, inputᵢ)
function transform!(lt::LossTransform)
lt.output.val[1] = sumvalue(lt.loss, value(lt.target), input_value(lt))
lt.output.val[1] = value(lt.loss, value(lt.target), input_value(lt), AvgMode.Sum())
Copy link
Contributor Author

@rdeits rdeits May 5, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I changed this based on JuliaML/LossFunctions.jl@774014d#diff-f2fc3674e12bf71a707cb4988076d39eR9

lt
end

Expand All @@ -76,12 +77,12 @@ immutable CrossEntropy{T} <: AbstractLossTransform{T}
target::SumNode{T,1}
output::OutputNode{T,1}

function CrossEntropy(n::Int)
function (::Type{CrossEntropy{T}}){T}(n::Int)
input = InputNode(T, n)
target = InputNode(T, n)
output = OutputNode(T, 1)
grad(output)[:] = one(T) # ∂L/∂L == 1
new(n, input, target, output)
new{T}(n, input, target, output)
end
end
CrossEntropy(n::Int) = CrossEntropy{Float64}(n)
Expand Down
2 changes: 1 addition & 1 deletion test/runtests.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,7 +6,7 @@ using Base.Test
t = Affine(nin, nout)
l = L2DistLoss()
λ = 1e-3
p = L2Penalty(λ)
p = scaled(L2Penalty(), λ)
obj = objective(t, l, p)
@show obj typeof(obj)

Expand Down