Slightly baffled why klDivergence : mixed x mixed -> float is busted.
Value: [1e-4 to 3e-2] hypothesis: I need to reintroduce `combineAlongSupportOfSecondArgument`
This commit is contained in:
Quinn Dougherty
parent
f2460a5e12
commit
249f58d3d8
|
|
@ -57,7 +57,7 @@ describe("klDivergence: continuous -> continuous -> float", () => {
|
|||
let kl = E.R.liftJoin2(klDivergence, prediction, answer)
|
||||
|
||||
switch kl {
|
||||
| Ok(kl') => kl'->expect->toBeSoCloseTo(analyticalKl, ~digits=3)
|
||||
| Ok(kl') => kl'->expect->toBeSoCloseTo(analyticalKl, ~digits=2)
|
||||
| Error(err) => {
|
||||
Js.Console.log(DistributionTypes.Error.toString(err))
|
||||
raise(KlFailed)
|
||||
|
|
|
|||
|
|
@ -155,6 +155,7 @@ module Score = {
|
|||
->PointSetDist_Scoring.DistEstimateDistAnswer
|
||||
->Ok
|
||||
)
|
||||
| (Score_Dist(_), _, Some(Ok(S(_)))) => DistributionTypes.Unreachable->Error
|
||||
| (Score_Dist(esti'), Score_Scalar(answ'), None) =>
|
||||
toPointSetFn(esti', ())->E.R.bind(esti'' =>
|
||||
{estimate: esti'', answer: answ', prior: None}
|
||||
|
|
@ -179,6 +180,7 @@ module Score = {
|
|||
->PointSetDist_Scoring.ScalarEstimateDistAnswer
|
||||
->Ok
|
||||
)
|
||||
| (Score_Scalar(_), _, Some(Ok(D(_)))) => DistributionTypes.Unreachable->Error
|
||||
| (Score_Scalar(esti'), Score_Scalar(answ'), None) =>
|
||||
{estimate: esti', answer: answ', prior: None}
|
||||
->PointSetDist_Scoring.ScalarEstimateScalarAnswer
|
||||
|
|
@ -199,44 +201,6 @@ module Score = {
|
|||
argsMake(~esti=estimate, ~answ=answer, ~prior)->E.R.bind(x =>
|
||||
x->PointSetDist.logScore->E.R2.errMap(y => DistributionTypes.OperationError(y))
|
||||
)
|
||||
|
||||
// let klDivergence = (prediction, answer, ~toPointSetFn: toPointSetFn): result<float, error> => {
|
||||
// let pointSets = E.R.merge(toPointSetFn(prediction), toPointSetFn(answer))
|
||||
// pointSets |> E.R2.bind(((predi, ans)) =>
|
||||
// PointSetDist.T.klDivergence(predi, ans)->E.R2.errMap(x => DistributionTypes.OperationError(x))
|
||||
// )
|
||||
// }
|
||||
//
|
||||
// let logScoreWithPointResolution = (
|
||||
// ~prediction: DistributionTypes.genericDist,
|
||||
// ~answer: float,
|
||||
// ~prior: option<DistributionTypes.genericDist>,
|
||||
// ~toPointSetFn: toPointSetFn,
|
||||
// ): result<float, error> => {
|
||||
// switch prior {
|
||||
// | Some(prior') =>
|
||||
// E.R.merge(toPointSetFn(prior'), toPointSetFn(prediction))->E.R.bind(((
|
||||
// prior'',
|
||||
// prediction'',
|
||||
// )) =>
|
||||
// PointSetDist.T.logScoreWithPointResolution(
|
||||
// ~prediction=prediction'',
|
||||
// ~answer,
|
||||
// ~prior=prior''->Some,
|
||||
// )->E.R2.errMap(x => DistributionTypes.OperationError(x))
|
||||
// )
|
||||
// | None =>
|
||||
// prediction
|
||||
// ->toPointSetFn
|
||||
// ->E.R.bind(x =>
|
||||
// PointSetDist.T.logScoreWithPointResolution(
|
||||
// ~prediction=x,
|
||||
// ~answer,
|
||||
// ~prior=None,
|
||||
// )->E.R2.errMap(x => DistributionTypes.OperationError(x))
|
||||
// )
|
||||
// }
|
||||
// }
|
||||
}
|
||||
/*
|
||||
PointSetDist.toSparkline calls "downsampleEquallyOverX", which downsamples it to n=bucketCount.
|
||||
|
|
|
|||
|
|
@ -270,20 +270,6 @@ module T = Dist({
|
|||
}
|
||||
let variance = (t: t): float =>
|
||||
XYShape.Analysis.getVarianceDangerously(t, mean, Analysis.getMeanOfSquares)
|
||||
|
||||
// let klDivergence = (prediction: t, answer: t) => {
|
||||
// let newShape = XYShape.PointwiseCombination.combineAlongSupportOfSecondArgument(
|
||||
// PointSetDist_Scoring.KLDivergence.integrand,
|
||||
// prediction.xyShape,
|
||||
// answer.xyShape,
|
||||
// )
|
||||
// newShape->E.R2.fmap(x => x->make->integralEndY)
|
||||
// }
|
||||
// let logScoreWithPointResolution = (~prediction: t, ~answer: float, ~prior: option<t>) => {
|
||||
// let priorPdf = prior->E.O2.fmap((shape, x) => XYShape.XtoY.linear(x, shape.xyShape))
|
||||
// let predictionPdf = x => XYShape.XtoY.linear(x, prediction.xyShape)
|
||||
// PointSetDist_Scoring.LogScoreWithPointResolution.score(~priorPdf, ~predictionPdf, ~answer)
|
||||
// }
|
||||
})
|
||||
|
||||
let isNormalized = (t: t): bool => {
|
||||
|
|
|
|||
|
|
@ -33,12 +33,6 @@ module type dist = {
|
|||
|
||||
let mean: t => float
|
||||
let variance: t => float
|
||||
// let klDivergence: (t, t) => result<float, Operation.Error.t>
|
||||
// let logScoreWithPointResolution: (
|
||||
// ~prediction: t,
|
||||
// ~answer: float,
|
||||
// ~prior: option<t>,
|
||||
// ) => result<float, Operation.Error.t>
|
||||
}
|
||||
|
||||
module Dist = (T: dist) => {
|
||||
|
|
@ -61,9 +55,6 @@ module Dist = (T: dist) => {
|
|||
let mean = T.mean
|
||||
let variance = T.variance
|
||||
let integralEndY = T.integralEndY
|
||||
// let klDivergence = T.klDivergence
|
||||
// let logScoreWithPointResolution = T.logScoreWithPointResolution
|
||||
|
||||
let updateIntegralCache = T.updateIntegralCache
|
||||
|
||||
module Integral = {
|
||||
|
|
|
|||
|
|
@ -195,27 +195,10 @@ module T = Dist({
|
|||
| Discrete(m) => Discrete.T.variance(m)
|
||||
| Continuous(m) => Continuous.T.variance(m)
|
||||
}
|
||||
|
||||
// let klDivergence = (prediction: t, answer: t) =>
|
||||
// switch (prediction, answer) {
|
||||
// | (Continuous(t1), Continuous(t2)) => Continuous.T.klDivergence(t1, t2)
|
||||
// | (Discrete(t1), Discrete(t2)) => Discrete.T.klDivergence(t1, t2)
|
||||
// | (m1, m2) => Mixed.T.klDivergence(m1->toMixed, m2->toMixed)
|
||||
// }
|
||||
//
|
||||
// let logScoreWithPointResolution = (~prediction: t, ~answer: float, ~prior: option<t>) => {
|
||||
// switch (prior, prediction) {
|
||||
// | (Some(Continuous(t1)), Continuous(t2)) =>
|
||||
// Continuous.T.logScoreWithPointResolution(~prediction=t2, ~answer, ~prior=t1->Some)
|
||||
// | (None, Continuous(t2)) =>
|
||||
// Continuous.T.logScoreWithPointResolution(~prediction=t2, ~answer, ~prior=None)
|
||||
// | _ => Error(Operation.NotYetImplemented)
|
||||
// }
|
||||
// }
|
||||
})
|
||||
|
||||
let logScore = (args: PointSetDist_Scoring.scoreArgs): result<float, Operation.Error.t> =>
|
||||
PointSetDist_Scoring.logScore(args, ~combineFn=combinePointwise, ~integrateFn=T.integralEndY)
|
||||
PointSetDist_Scoring.logScore(args, ~combineFn=combinePointwise, ~integrateFn=T.Integral.sum)
|
||||
|
||||
let pdf = (f: float, t: t) => {
|
||||
let mixedPoint: PointSetTypes.mixedPoint = T.xToY(f, t)
|
||||
|
|
|
|||
|
|
@ -1,7 +1,4 @@
|
|||
type t = PointSetTypes.pointSetDist
|
||||
type continuousShape = PointSetTypes.continuousShape
|
||||
type discreteShape = PointSetTypes.discreteShape
|
||||
type mixedShape = PointSetTypes.mixedShape
|
||||
|
||||
type scalar = float
|
||||
type abstractScoreArgs<'a, 'b> = {estimate: 'a, answer: 'b, prior: option<'a>}
|
||||
|
|
@ -71,14 +68,14 @@ module WithScalarAnswer = {
|
|||
minusScaledLogOfQuot(~esti=numerator, ~answ=priorDensityOfAnswer)
|
||||
}
|
||||
}
|
||||
|
||||
let sum = (mp: PointSetTypes.MixedPoint.t): float => mp.continuous +. mp.discrete
|
||||
let score = (~estimate: t, ~answer: scalar): result<float, Operation.Error.t> => {
|
||||
let estimatePdf = x =>
|
||||
switch estimate {
|
||||
| Continuous(esti) => XYShape.XtoY.linear(x, esti.xyShape)
|
||||
| Discrete(esti) => XYShape.XtoY.linear(x, esti.xyShape)
|
||||
| Mixed(esti) =>
|
||||
XYShape.XtoY.linear(x, esti.continuous.xyShape) +.
|
||||
XYShape.XtoY.linear(x, esti.discrete.xyShape)
|
||||
| Continuous(esti) => Continuous.T.xToY(x, esti)->sum
|
||||
| Discrete(esti) => Discrete.T.xToY(x, esti)->sum
|
||||
| Mixed(esti) => Mixed.T.xToY(x, esti)->sum
|
||||
}
|
||||
|
||||
score'(~estimatePdf, ~answer)
|
||||
|
|
@ -89,19 +86,15 @@ module WithScalarAnswer = {
|
|||
> => {
|
||||
let estimatePdf = x =>
|
||||
switch estimate {
|
||||
| Continuous(esti) => XYShape.XtoY.linear(x, esti.xyShape)
|
||||
| Discrete(esti) => XYShape.XtoY.linear(x, esti.xyShape)
|
||||
| Mixed(esti) =>
|
||||
XYShape.XtoY.linear(x, esti.continuous.xyShape) +.
|
||||
XYShape.XtoY.linear(x, esti.discrete.xyShape)
|
||||
| Continuous(esti) => Continuous.T.xToY(x, esti)->sum
|
||||
| Discrete(esti) => Discrete.T.xToY(x, esti)->sum
|
||||
| Mixed(esti) => Mixed.T.xToY(x, esti)->sum
|
||||
}
|
||||
let priorPdf = x =>
|
||||
switch prior {
|
||||
| Continuous(prio) => XYShape.XtoY.linear(x, prio.xyShape)
|
||||
| Discrete(prio) => XYShape.XtoY.linear(x, prio.xyShape)
|
||||
| Mixed(prio) =>
|
||||
XYShape.XtoY.linear(x, prio.continuous.xyShape) +.
|
||||
XYShape.XtoY.linear(x, prio.discrete.xyShape)
|
||||
| Continuous(prio) => Continuous.T.xToY(x, prio)->sum
|
||||
| Discrete(prio) => Discrete.T.xToY(x, prio)->sum
|
||||
| Mixed(prio) => Mixed.T.xToY(x, prio)->sum
|
||||
}
|
||||
scoreWithPrior'(~estimatePdf, ~answer, ~priorPdf)
|
||||
}
|
||||
|
|
|
|||
|
|
@ -541,6 +541,7 @@ module A = {
|
|||
let init = Array.init
|
||||
let reduce = Belt.Array.reduce
|
||||
let reducei = Belt.Array.reduceWithIndex
|
||||
let some = Belt.Array.some
|
||||
let isEmpty = r => length(r) < 1
|
||||
let stableSortBy = Belt.SortArray.stableSortBy
|
||||
let toNoneIfEmpty = r => isEmpty(r) ? None : Some(r)
|
||||
|
|
|
|||
Loading…
Reference in New Issue
Block a user