Skip to content

Commit 3de0745

Browse files
lrnvlrnv
committed
up
1 parent 75dafd4 commit 3de0745

File tree

1 file changed

+85
-86
lines changed

1 file changed

+85
-86
lines changed

‎benchmark/benchmarks.jl‎

Lines changed: 85 additions & 86 deletions
Original file line numberDiff line numberDiff line change
@@ -2,41 +2,20 @@ using BenchmarkTools
22
using Copulas
33
using Distributions
44
using StableRNGs
5+
const rng = StableRNG(123)
56

67
# PkgBenchmark entrypoint; must define SUITE
78
const SUITE = BenchmarkGroup()
8-
module M
9-
using StableRNGs
10-
const rng = StableRNG(123)
11-
end
129

13-
# Separate SklarDist benchmarks on a single representative model (Clayton d=5)
14-
let rng = StableRNG(321)
15-
top = SUITE["sklar"] = BenchmarkGroup()
16-
C = ClaytonCopula(5, 2.0)
17-
m = (Distributions.Normal(), Distributions.LogNormal(), Distributions.Gamma(2,2), Distributions.Beta(2,5), Distributions.Uniform())
18-
S = SklarDist(C, m)
19-
X = rand(rng, S, 128)
20-
U = pseudos(X)
21-
top["rand/128"] = @benchmarkable rand($rng, $S, 128)
22-
top["cdf/128"] = @benchmarkable cdf($S, $X)
23-
top["pdf/128"] = @benchmarkable pdf($S, $X)
24-
top["rosenblatt/128"] = @benchmarkable rosenblatt($S, $X)
25-
V = rand(rng, 5, 128)
26-
top["inverse_rosenblatt/128"] = @benchmarkable inverse_rosenblatt($S, $V)
27-
# Fitting pathways (keep light)
28-
top["fit/ifm"] = @benchmarkable Distributions.fit(CopulaModel, SklarDist{ClaytonCopula, typeof(m)}, $X; sklar_method=:ifm, copula_method=:itau, summaries=false)
29-
top["fit/ecdf"] = @benchmarkable Distributions.fit(CopulaModel, SklarDist{ClaytonCopula, typeof(m)}, $X; sklar_method=:ecdf, copula_method=:itau, summaries=false)
30-
end
3110

3211
const EXAMPLES = unique([
3312
AMHCopula(2,-1.0),
34-
AMHCopula(2,-rand(M.rng)),
13+
AMHCopula(2,-rand(rng)),
3514
AMHCopula(2,0.7),
36-
AMHCopula(2,rand(M.rng)),
37-
AMHCopula(3,-rand(M.rng)*0.1),
15+
AMHCopula(2,rand(rng)),
16+
AMHCopula(3,-rand(rng)*0.1),
3817
AMHCopula(3,0.6),
39-
AMHCopula(3,rand(M.rng)),
18+
AMHCopula(3,rand(rng)),
4019
AMHCopula(4,-0.01),
4120
ArchimaxCopula(2, Copulas.BB1Generator(1.3, 1.4), Copulas.AsymGalambosTail(0.35, 0.65, 0.3)),
4221
ArchimaxCopula(2, Copulas.BB1Generator(1.3, 1.4), Copulas.GalambosTail(0.7)),
@@ -110,33 +89,33 @@ const EXAMPLES = unique([
11089
ArchimaxCopula(2, Copulas.JoeGenerator(2.5), Copulas.LogTail(2.0)),
11190
ArchimedeanCopula(10,i𝒲(Dirac(1),10)),
11291
ArchimedeanCopula(10,i𝒲(MixtureModel([Dirac(1), Dirac(2)]),11)),
113-
ArchimedeanCopula(2, EmpiricalGenerator(randn(M.rng, 4, 150))),
92+
ArchimedeanCopula(2, EmpiricalGenerator(randn(rng, 4, 150))),
11493
ArchimedeanCopula(2,i𝒲(LogNormal(),2)),
11594
ArchimedeanCopula(2,i𝒲(LogNormal(3),5)),
11695
ArchimedeanCopula(2,i𝒲(Pareto(1),5)),
117-
ArchimedeanCopula(3, EmpiricalGenerator(randn(M.rng, 3, 200))),
96+
ArchimedeanCopula(3, EmpiricalGenerator(randn(rng, 3, 200))),
11897
AsymGalambosCopula(2, 0.1, 0.2, 0.6),
11998
AsymGalambosCopula(2, 0.6129496106778634, 0.820474440393214, 0.22304578643880224),
120-
AsymGalambosCopula(2, 10+5*rand(M.rng), 1.0, 1.0),
121-
AsymGalambosCopula(2, 10+5*rand(M.rng), rand(M.rng), rand(M.rng)),
99+
AsymGalambosCopula(2, 10+5*rand(rng), 1.0, 1.0),
100+
AsymGalambosCopula(2, 10+5*rand(rng), rand(rng), rand(rng)),
122101
AsymGalambosCopula(2, 11.647356700032505, 0.6195348270893413, 0.4197760589260566),
123102
AsymGalambosCopula(2, 5.0, 0.8, 0.3),
124-
AsymGalambosCopula(2, 5+4*rand(M.rng), 1.0, 1.0),
125-
AsymGalambosCopula(2, 5+4*rand(M.rng), rand(M.rng), rand(M.rng)),
103+
AsymGalambosCopula(2, 5+4*rand(rng), 1.0, 1.0),
104+
AsymGalambosCopula(2, 5+4*rand(rng), rand(rng), rand(rng)),
126105
AsymGalambosCopula(2, 8.810168494949659, 0.5987759444612732, 0.5391280234619427),
127-
AsymGalambosCopula(2, rand(M.rng), 1.0, 1.0),
128-
AsymGalambosCopula(2, rand(M.rng), rand(M.rng), rand(M.rng)),
106+
AsymGalambosCopula(2, rand(rng), 1.0, 1.0),
107+
AsymGalambosCopula(2, rand(rng), rand(rng), rand(rng)),
129108
AsymLogCopula(2, 1.0, 0.0, 0.0),
130109
AsymLogCopula(2, 1.0, 1.0, 1.0),
131-
AsymLogCopula(2, 1.0, rand(M.rng), rand(M.rng)),
110+
AsymLogCopula(2, 1.0, rand(rng), rand(rng)),
132111
AsymLogCopula(2, 1.2, 0.3,0.6),
133112
AsymLogCopula(2, 1.5, 0.5, 0.2),
134-
AsymLogCopula(2, 1+4*rand(M.rng), 0.0, 0.0),
135-
AsymLogCopula(2, 1+4*rand(M.rng), 1.0, 1.0),
136-
AsymLogCopula(2, 1+4*rand(M.rng), rand(M.rng), rand(M.rng)),
137-
AsymLogCopula(2, 10+5*rand(M.rng), 0.0, 0.0),
138-
AsymLogCopula(2, 10+5*rand(M.rng), 1.0, 1.0),
139-
AsymLogCopula(2, 10+5*rand(M.rng), rand(M.rng), rand(M.rng)),
113+
AsymLogCopula(2, 1+4*rand(rng), 0.0, 0.0),
114+
AsymLogCopula(2, 1+4*rand(rng), 1.0, 1.0),
115+
AsymLogCopula(2, 1+4*rand(rng), rand(rng), rand(rng)),
116+
AsymLogCopula(2, 10+5*rand(rng), 0.0, 0.0),
117+
AsymLogCopula(2, 10+5*rand(rng), 1.0, 1.0),
118+
AsymLogCopula(2, 10+5*rand(rng), rand(rng), rand(rng)),
140119
AsymMixedCopula(2, 0.1, 0.2),
141120
AsymMixedCopula(2, 0.12, 0.13),
142121
BB10Copula(2, 1.5, 0.7),
@@ -169,27 +148,27 @@ const EXAMPLES = unique([
169148
BC2Copula(2, 0.7,0.3),
170149
BC2Copula(2, 1.0, 0.0),
171150
BC2Copula(2, 1/2,1/2),
172-
BC2Copula(2, rand(M.rng), rand(M.rng)),
151+
BC2Copula(2, rand(rng), rand(rng)),
173152
BernsteinCopula(ArchimaxCopula(2, Copulas.FrankGenerator(0.8), Copulas.HuslerReissTail(0.6)); m=5),
174153
BernsteinCopula(ClaytonCopula(3, 3.3); m=5),
175154
BernsteinCopula(GalambosCopula(2, 2.5); m=5),
176155
BernsteinCopula(GaussianCopula(2, 0.3); m=5),
177156
BernsteinCopula(IndependentCopula(4); m=5),
178157
BernsteinCopula(IndependentCopula(4); m=5),
179-
BernsteinCopula(randn(M.rng, 2,100), pseudo_values=false),
180-
BetaCopula(randn(M.rng, 2,100)),
181-
BetaCopula(randn(M.rng, 3,100)),
182-
CheckerboardCopula(randn(M.rng, 2,100); pseudo_values=false),
183-
CheckerboardCopula(randn(M.rng, 3,100); pseudo_values=false),
184-
CheckerboardCopula(randn(M.rng, 4,100); pseudo_values=false),
158+
BernsteinCopula(randn(rng, 2,100), pseudo_values=false),
159+
BetaCopula(randn(rng, 2,100)),
160+
BetaCopula(randn(rng, 3,100)),
161+
CheckerboardCopula(randn(rng, 2,100); pseudo_values=false),
162+
CheckerboardCopula(randn(rng, 3,100); pseudo_values=false),
163+
CheckerboardCopula(randn(rng, 4,100); pseudo_values=false),
185164
ClaytonCopula(2,-0.7),
186-
ClaytonCopula(2,-log(rand(M.rng))),
187-
ClaytonCopula(2,-rand(M.rng)),
165+
ClaytonCopula(2,-log(rand(rng))),
166+
ClaytonCopula(2,-rand(rng)),
188167
ClaytonCopula(2,7),
189-
ClaytonCopula(3,-log(rand(M.rng))),
190-
ClaytonCopula(3,-rand(M.rng)/2),
191-
ClaytonCopula(4,-log(rand(M.rng))),
192-
ClaytonCopula(4,-rand(M.rng)/3),
168+
ClaytonCopula(3,-log(rand(rng))),
169+
ClaytonCopula(3,-rand(rng)/2),
170+
ClaytonCopula(4,-log(rand(rng))),
171+
ClaytonCopula(4,-rand(rng)/3),
193172
ClaytonCopula(4,7.),
194173
Copulas.SubsetCopula(RafteryCopula(3, 0.5), (2,1)),
195174
CuadrasAugeCopula(2, 0.0),
@@ -198,54 +177,54 @@ const EXAMPLES = unique([
198177
CuadrasAugeCopula(2, 0.7103550345192344),
199178
CuadrasAugeCopula(2, 0.8),
200179
CuadrasAugeCopula(2, 1.0),
201-
CuadrasAugeCopula(2, rand(M.rng)),
180+
CuadrasAugeCopula(2, rand(rng)),
202181
EmpiricalCopula(randn(2,10),pseudo_values=false),
203182
EmpiricalCopula(randn(2,20),pseudo_values=false),
204-
EmpiricalEVCopula(randn(M.rng, 2,10); method=:cfg, pseudo_values=false),
205-
EmpiricalEVCopula(randn(M.rng, 2,10); method=:ols, pseudo_values=false),
206-
EmpiricalEVCopula(randn(M.rng, 2,10); method=:pickands, pseudo_values=false),
207-
EmpiricalEVCopula(randn(M.rng, 2,20); method=:cfg, pseudo_values=false),
208-
EmpiricalEVCopula(randn(M.rng, 2,20); method=:ols, pseudo_values=false),
209-
EmpiricalEVCopula(randn(M.rng, 2,20); method=:pickands, pseudo_values=false),
183+
EmpiricalEVCopula(randn(rng, 2,10); method=:cfg, pseudo_values=false),
184+
EmpiricalEVCopula(randn(rng, 2,10); method=:ols, pseudo_values=false),
185+
EmpiricalEVCopula(randn(rng, 2,10); method=:pickands, pseudo_values=false),
186+
EmpiricalEVCopula(randn(rng, 2,20); method=:cfg, pseudo_values=false),
187+
EmpiricalEVCopula(randn(rng, 2,20); method=:ols, pseudo_values=false),
188+
EmpiricalEVCopula(randn(rng, 2,20); method=:pickands, pseudo_values=false),
210189
FGMCopula(2, 0.0),
211-
FGMCopula(2, rand(M.rng)),
190+
FGMCopula(2, rand(rng)),
212191
FGMCopula(2,1),
213192
FGMCopula(3, [0.3,0.3,0.3,0.3]),
214193
FGMCopula(3,[0.1,0.2,0.3,0.4]),
215194
FrankCopula(2,-5),
216195
FrankCopula(2,0.5),
217-
FrankCopula(2,1-log(rand(M.rng))),
196+
FrankCopula(2,1-log(rand(rng))),
218197
FrankCopula(2,1.0),
219-
FrankCopula(3,1-log(rand(M.rng))),
198+
FrankCopula(3,1-log(rand(rng))),
220199
FrankCopula(3,1.0),
221200
FrankCopula(3,12),
222-
FrankCopula(4,1-log(rand(M.rng))),
201+
FrankCopula(4,1-log(rand(rng))),
223202
FrankCopula(4,1.0),
224203
FrankCopula(4,150),
225204
FrankCopula(4,30),
226205
FrankCopula(4,37),
227206
GalambosCopula(2, 0.3),
228-
GalambosCopula(2, 1+4*rand(M.rng)),
207+
GalambosCopula(2, 1+4*rand(rng)),
229208
GalambosCopula(2, 120),
230209
GalambosCopula(2, 20),
231210
GalambosCopula(2, 210),
232211
GalambosCopula(2, 4.3),
233-
GalambosCopula(2, 5+5*rand(M.rng)),
212+
GalambosCopula(2, 5+5*rand(rng)),
234213
GalambosCopula(2, 80),
235-
GalambosCopula(2, rand(M.rng)),
214+
GalambosCopula(2, rand(rng)),
236215
GaussianCopula([1 0.5; 0.5 1]),
237216
GaussianCopula([1 0.7; 0.7 1]),
238217
GumbelBarnettCopula(2,1.0),
239-
GumbelBarnettCopula(2,rand(M.rng)),
218+
GumbelBarnettCopula(2,rand(rng)),
240219
GumbelBarnettCopula(3,0.1),
241220
GumbelBarnettCopula(3,0.35),
242-
GumbelBarnettCopula(3,rand(M.rng)*0.38),
221+
GumbelBarnettCopula(3,rand(rng)*0.38),
243222
GumbelBarnettCopula(4,0.2),
244223
GumbelCopula(2, 1.2),
245-
GumbelCopula(2,1-log(rand(M.rng))),
224+
GumbelCopula(2,1-log(rand(rng))),
246225
GumbelCopula(2,8),
247-
GumbelCopula(3,1-log(rand(M.rng))),
248-
GumbelCopula(4,1-log(rand(M.rng))),
226+
GumbelCopula(3,1-log(rand(rng))),
227+
GumbelCopula(4,1-log(rand(rng))),
249228
GumbelCopula(4,100),
250229
GumbelCopula(4,20),
251230
GumbelCopula(4,7),
@@ -256,22 +235,22 @@ const EXAMPLES = unique([
256235
HuslerReissCopula(2, 5.319851350643586),
257236
IndependentCopula(2),
258237
IndependentCopula(3),
259-
InvGaussianCopula(2,-log(rand(M.rng))),
238+
InvGaussianCopula(2,-log(rand(rng))),
260239
InvGaussianCopula(2,1.0),
261-
InvGaussianCopula(2,rand(M.rng)),
262-
InvGaussianCopula(3,-log(rand(M.rng))),
263-
InvGaussianCopula(3,rand(M.rng)),
264-
InvGaussianCopula(4,-log(rand(M.rng))),
240+
InvGaussianCopula(2,rand(rng)),
241+
InvGaussianCopula(3,-log(rand(rng))),
242+
InvGaussianCopula(3,rand(rng)),
243+
InvGaussianCopula(4,-log(rand(rng))),
265244
InvGaussianCopula(4,0.05),
266245
InvGaussianCopula(4,1.0),
267-
JoeCopula(2,1-log(rand(M.rng))),
246+
JoeCopula(2,1-log(rand(rng))),
268247
JoeCopula(2,3),
269248
JoeCopula(2,Inf),
270-
JoeCopula(3,1-log(rand(M.rng))),
249+
JoeCopula(3,1-log(rand(rng))),
271250
JoeCopula(3,7),
272-
JoeCopula(4,1-log(rand(M.rng))),
251+
JoeCopula(4,1-log(rand(rng))),
273252
LogCopula(2, 1.5),
274-
LogCopula(2, 1+9*rand(M.rng)),
253+
LogCopula(2, 1+9*rand(rng)),
275254
LogCopula(2, 5.5),
276255
MCopula(2),
277256
MCopula(4),
@@ -283,7 +262,7 @@ const EXAMPLES = unique([
283262
MOCopula(2, 0.5, 0.5, 0.5),
284263
MOCopula(2, 0.5960710257852946, 0.3313524247810329, 0.09653466861970061),
285264
MOCopula(2, 1.0, 1.0, 1.0),
286-
MOCopula(2, rand(M.rng), rand(M.rng), rand(M.rng)),
265+
MOCopula(2, rand(rng), rand(rng), rand(rng)),
287266
PlackettCopula(0.5),
288267
PlackettCopula(0.8),
289268
PlackettCopula(2.0),
@@ -298,18 +277,18 @@ const EXAMPLES = unique([
298277
tEVCopula(2, 2.0, 0.5),
299278
tEVCopula(2, 3.0, 0.0),
300279
tEVCopula(2, 4.0, 0.5),
301-
tEVCopula(2, 4+6*rand(M.rng), -0.9+1.9*rand(M.rng)),
280+
tEVCopula(2, 4+6*rand(rng), -0.9+1.9*rand(rng)),
302281
tEVCopula(2, 5.0, -0.5),
303282
tEVCopula(2, 5.466564460573727, -0.6566645244416698),
304283
WCopula(2),
305-
])
284+
]);
306285

307286
let rng = StableRNG(123)
308287
top = SUITE["copulas"] = BenchmarkGroup()
309288
# helper to get stable display names
310289
_show_name(C) = sprint(show, C)
311290
_base_from_show(s::AbstractString) = (m = match(r"^[^\s(]+", s); m === nothing ? s : m.match)
312-
for C in EXAMPLES
291+
for C in EXAMPLES[1:2] # only the first two for the moment to see.
313292
CT, d = typeof(C), length(C)
314293
name_full = _show_name(C)
315294
name_base = _base_from_show(name_full)
@@ -363,3 +342,23 @@ let rng = StableRNG(123)
363342
end
364343
end
365344
end
345+
346+
347+
# Separate SklarDist benchmarks on a single representative model (Clayton d=5)
348+
# let rng = StableRNG(321)
349+
# top = SUITE["sklar"] = BenchmarkGroup()
350+
# C = ClaytonCopula(5, 2.0)
351+
# m = (Distributions.Normal(), Distributions.LogNormal(), Distributions.Gamma(2,2), Distributions.Beta(2,5), Distributions.Uniform())
352+
# S = SklarDist(C, m)
353+
# X = rand(rng, S, 128)
354+
# U = pseudos(X)
355+
# top["rand/128"] = @benchmarkable rand($rng, $S, 128)
356+
# top["cdf/128"] = @benchmarkable cdf($S, $X)
357+
# top["pdf/128"] = @benchmarkable pdf($S, $X)
358+
# top["rosenblatt/128"] = @benchmarkable rosenblatt($S, $X)
359+
# V = rand(rng, 5, 128)
360+
# top["inverse_rosenblatt/128"] = @benchmarkable inverse_rosenblatt($S, $V)
361+
# # Fitting pathways (keep light)
362+
# top["fit/ifm"] = @benchmarkable Distributions.fit(CopulaModel, SklarDist{ClaytonCopula, typeof(m)}, $X; sklar_method=:ifm, copula_method=:itau, summaries=false)
363+
# top["fit/ecdf"] = @benchmarkable Distributions.fit(CopulaModel, SklarDist{ClaytonCopula, typeof(m)}, $X; sklar_method=:ecdf, copula_method=:itau, summaries=false)
364+
# end

0 commit comments

Comments
 (0)