Add compatibility with LossFunctions.jl, test with least squares

Project.toml

@@ -7,8 +7,16 @@ version = "0.1.0"
 ForwardDiff = "f6369f11-7733-5829-9624-2563aa707210"
 LBFGSB = "5be7bae1-8223-5378-bac3-9e7378a2f6e6"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
+LossFunctions = "30fc2ffe-d236-52d8-8643-a9d8f7c094a7"
 
 [compat]
 ForwardDiff = "0.10.36"
 LBFGSB = "0.4.1"
+LossFunctions = "0.11.1"
 julia = "1.6"
+
+[extensions]
+LossFunctionsExt = "LossFunctions"
+
+[weakdeps]
+LossFunctions = "30fc2ffe-d236-52d8-8643-a9d8f7c094a7"

ext/LossFunctionsExt.jl

@@ -0,0 +1,20 @@
+module LossFunctionsExt
+
+using GCPDecompositions, LossFunctions
+
+"""
+    gcp(X::Array, r, loss::LossFunctions.SupervisedLoss, lower]) -> CPD
+
+Compute an approximate rank-`r` CP decomposition of the tensor `X`
+with respect to the loss function `loss` and return a `CPD` object.
+
+# Inputs
++ `X` : multi-dimensional tensor/array to approximate/decompose
++ `r` : number of components for the CPD
++ `loss` : loss function from LossFunctions.jl
++ `lower` : lower bound for factor matrix entries, `default = -Inf`
+"""
+GCPDecompositions.gcp(X::Array, r, loss::LossFunctions.SupervisedLoss, lower=-Inf) =
+    GCPDecompositions._gcp(X, r, (x, m) -> loss(m, x), (x, m) -> LossFunctions.deriv(loss, m, x), lower, (;))
+
+end

src/GCPDecompositions.jl

@@ -17,4 +17,8 @@ export gcp
 include("type-cpd.jl")
 include("gcp-opt.jl")
 
+if !isdefined(Base, :get_extension)
+    include("../ext/LossFunctionsExt.jl")
+end
+
 end

test/Manifest.toml

@@ -2,7 +2,7 @@
 
 julia_version = "1.8.2"
 manifest_format = "2.0"
-project_hash = "3b0347c951b114bc9938540028de5bc8d1fc29c5"
+project_hash = "5b46ab02fc1fd08dc4db50c5af3027ccd5dc095b"
 
 [[deps.Adapt]]
 deps = ["LinearAlgebra"]
@@ -39,6 +39,12 @@ uuid = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
 [[deps.Logging]]
 uuid = "56ddb016-857b-54e1-b83d-db4d58db5568"
 
+[[deps.LossFunctions]]
+deps = ["Markdown", "Requires", "Statistics"]
+git-tree-sha1 = "df9da07efb9b05ca7ef701acec891ee8f73c99e2"
+uuid = "30fc2ffe-d236-52d8-8643-a9d8f7c094a7"
+version = "0.11.1"
+
 [[deps.Markdown]]
 deps = ["Base64"]
 uuid = "d6f4376e-aef5-505a-96c1-9c027394607a"
@@ -62,13 +68,27 @@ uuid = "de0858da-6303-5e67-8744-51eddeeeb8d7"
 deps = ["SHA", "Serialization"]
 uuid = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
 
+[[deps.Requires]]
+deps = ["UUIDs"]
+git-tree-sha1 = "838a3a4188e2ded87a4f9f184b4b0d78a1e91cb7"
+uuid = "ae029012-a4dd-5104-9daa-d747884805df"
+version = "1.3.0"
+
 [[deps.SHA]]
 uuid = "ea8e919c-243c-51af-8825-aaa63cd721ce"
 version = "0.7.0"
 
 [[deps.Serialization]]
 uuid = "9e88b42a-f829-5b0c-bbe9-9e923198166b"
 
+[[deps.SparseArrays]]
+deps = ["LinearAlgebra", "Random"]
+uuid = "2f01184e-e22b-5df5-ae63-d93ebab69eaf"
+
+[[deps.Statistics]]
+deps = ["LinearAlgebra", "SparseArrays"]
+uuid = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
+
 [[deps.TOML]]
 deps = ["Dates"]
 uuid = "fa267f1f-6049-4f14-aa54-33bafae1ed76"
@@ -89,6 +109,10 @@ git-tree-sha1 = "8621ba2637b49748e2dc43ba3d84340be2938022"
 uuid = "1c621080-faea-4a02-84b6-bbd5e436b8fe"
 version = "0.1.1"
 
+[[deps.UUIDs]]
+deps = ["Random", "SHA"]
+uuid = "cf7118a7-6976-5b1a-9a39-7adc72f591a4"
+
 [[deps.Unicode]]
 uuid = "4ec0a83e-493e-50e2-b9ac-8f72acf5a8f5"
 

test/Project.toml

@@ -1,9 +1,11 @@
 [deps]
+LossFunctions = "30fc2ffe-d236-52d8-8643-a9d8f7c094a7"
 OffsetArrays = "6fe1bfb0-de20-5000-8ca7-80f57d26f881"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 TestItemRunner = "f8b46487-2199-4994-9208-9a1283c18c0a"
 
 [compat]
+LossFunctions = "0.11.1"
 OffsetArrays = "1.12.7"
 TestItemRunner = "0.2.1"

test/items/gcp-opt.jl

@@ -2,6 +2,7 @@
 
 @testitem "least squares" begin
     using Random
+    using LossFunctions
 
     @testset "size(X)=$sz, rank(X)=$r" for sz in [(15, 20, 25), (30, 40, 50)], r in 1:2
         Random.seed!(0)
@@ -17,5 +18,8 @@
 
         Mh = gcp(X, r) # test default (least-squares) loss
         @test maximum(I -> abs(Mh[I] - X[I]), CartesianIndices(X)) <= 1e-5
+
+        Mh = gcp(X, r, L2DistLoss()) # test with loss function from LossFunctions.jl (least squares)
+        @test maximum(I -> abs(Mh[I] - X[I]), CartesianIndices(X)) <= 1e-5
     end
 end