squiggle/packages/squiggle-lang/__tests__/Distributions/Mixture_test.res

71 lines
2.9 KiB
Plaintext
Raw Permalink Normal View History

2022-04-07 02:24:00 +00:00
open Jest
open Expect
2022-04-07 22:38:49 +00:00
open TestHelpers
2022-04-07 02:24:00 +00:00
2022-04-07 22:38:49 +00:00
// TODO: use Normal.make (etc.), but preferably after the new validation dispatch is in.
2022-04-07 02:24:00 +00:00
let mkNormal = (mean, stdev) => GenericDist_Types.Symbolic(#Normal({mean: mean, stdev: stdev}))
let mkBeta = (alpha, beta) => GenericDist_Types.Symbolic(#Beta({alpha: alpha, beta: beta}))
let mkExponential = rate => GenericDist_Types.Symbolic(#Exponential({rate: rate}))
2022-04-07 17:33:12 +00:00
let mkUniform = (low, high) => GenericDist_Types.Symbolic(#Uniform({low: low, high: high}))
let mkCauchy = (local, scale) => GenericDist_Types.Symbolic(#Cauchy({local: local, scale: scale}))
let mkLognormal = (mu, sigma) => GenericDist_Types.Symbolic(#Lognormal({mu: mu, sigma: sigma}))
2022-04-07 02:24:00 +00:00
describe("mixture", () => {
testAll("fair mean of two normal distributions", list{(0.0, 1e2), (-1e1, -1e-4), (-1e1, 1e2), (-1e1, 1e1)}, tup => { // should be property
let (mean1, mean2) = tup
2022-04-07 22:38:49 +00:00
let meanValue = {
2022-04-07 02:24:00 +00:00
run(Mixture([(mkNormal(mean1, 9e-1), 0.5), (mkNormal(mean2, 9e-1), 0.5)]))
-> outputMap(FromDist(ToFloat(#Mean)))
}
2022-04-07 22:38:49 +00:00
meanValue -> unpackFloat -> expect -> toBeSoCloseTo((mean1 +. mean2) /. 2.0, ~digits=-1)
2022-04-07 02:24:00 +00:00
})
testAll(
"weighted mean of a beta and an exponential",
// This would not survive property testing, it was easy for me to find cases that NaN'd out.
list{((128.0, 1.0), 2.0), ((2e-1, 64.0), 16.0), ((1e0, 1e0), 64.0)},
tup => {
2022-04-07 22:38:49 +00:00
let ((alpha, beta), rate) = tup
let betaWeight = 0.25
let exponentialWeight = 0.75
let meanValue = {
2022-04-07 12:55:52 +00:00
run(Mixture(
[
2022-04-07 22:38:49 +00:00
(mkBeta(alpha, beta), betaWeight),
(mkExponential(rate), exponentialWeight)
2022-04-07 12:55:52 +00:00
]
)) -> outputMap(FromDist(ToFloat(#Mean)))
2022-04-07 02:24:00 +00:00
}
2022-04-07 22:38:49 +00:00
let betaMean = 1.0 /. (1.0 +. beta /. alpha)
let exponentialMean = 1.0 /. rate
meanValue
2022-04-07 02:24:00 +00:00
-> unpackFloat
-> expect
-> toBeSoCloseTo(
2022-04-07 22:38:49 +00:00
betaWeight *. betaMean +. exponentialWeight *. exponentialMean,
2022-04-07 02:24:00 +00:00
~digits=-1
)
}
)
2022-04-07 17:33:12 +00:00
testAll(
"weighted mean of lognormal and uniform",
// Would not survive property tests: very easy to find cases that NaN out.
list{((-1e2,1e1), (2e0,1e0)), ((-1e-16,1e-16), (1e-8,1e0)), ((0.0,1e0), (1e0,1e-2))},
2022-04-07 17:33:12 +00:00
tup => {
2022-04-07 22:38:49 +00:00
let ((low, high), (mu, sigma)) = tup
let uniformWeight = 0.6
let lognormalWeight = 0.4
let meanValue = {
run(Mixture([(mkUniform(low, high), uniformWeight), (mkLognormal(mu, sigma), lognormalWeight)]))
2022-04-07 17:33:12 +00:00
-> outputMap(FromDist(ToFloat(#Mean)))
}
2022-04-07 22:38:49 +00:00
let uniformMean = (low +. high) /. 2.0
let lognormalMean = mu +. sigma ** 2.0 /. 2.0
meanValue
2022-04-07 17:33:12 +00:00
-> unpackFloat
-> expect
2022-04-07 22:38:49 +00:00
-> toBeSoCloseTo(uniformWeight *. uniformMean +. lognormalWeight *. lognormalMean, ~digits=-1)
2022-04-07 17:33:12 +00:00
}
)
2022-04-07 02:24:00 +00:00
})