293 lines
8.2 KiB
ReasonML
293 lines
8.2 KiB
ReasonML
open DistributionTypes;
|
|
|
|
let _lastElement = (a: array('a)) =>
|
|
switch (Belt.Array.size(a)) {
|
|
| 0 => None
|
|
| n => Belt.Array.get(a, n - 1)
|
|
};
|
|
|
|
type pointInRange =
|
|
| Unbounded
|
|
| X(float);
|
|
|
|
module XYShape = {
|
|
type t = xyShape;
|
|
|
|
let toJs = (t: t) => {
|
|
{"xs": t.xs, "ys": t.ys};
|
|
};
|
|
|
|
let minX = (t: t) => t.xs |> E.A.get(_, 0);
|
|
// TODO: Check if this actually gets the last element, I'm not sure it does.
|
|
let maxX = (t: t) => t.xs |> (r => E.A.get(r, E.A.length(r) - 1));
|
|
|
|
let zip = t => Belt.Array.zip(t.xs, t.ys);
|
|
|
|
let fmap = (t: t, y): t => {xs: t.xs, ys: t.ys |> E.A.fmap(y)};
|
|
|
|
let scaleCdfTo = (~scaleTo=1., t: t) =>
|
|
switch (_lastElement(t.ys)) {
|
|
| Some(n) =>
|
|
let scaleBy = scaleTo /. n;
|
|
fmap(t, r => r *. scaleBy);
|
|
| None => t
|
|
};
|
|
|
|
let yFold = (fn, t: t) => {
|
|
E.A.fold_left(fn, 0., t.ys);
|
|
};
|
|
|
|
let ySum = yFold((a, b) => a +. b);
|
|
|
|
let fromArray = ((xs, ys)): t => {xs, ys};
|
|
let fromArrays = (xs, ys): t => {xs, ys};
|
|
|
|
let _transverse = fn =>
|
|
Belt.Array.reduce(_, [||], (items, (x, y)) =>
|
|
switch (_lastElement(items)) {
|
|
| Some((xLast, yLast)) =>
|
|
Belt.Array.concat(items, [|(x, fn(y, yLast))|])
|
|
| None => [|(x, y)|]
|
|
}
|
|
);
|
|
|
|
let _transverseShape = (fn, p: t) => {
|
|
Belt.Array.zip(p.xs, p.ys)
|
|
|> _transverse(fn)
|
|
|> Belt.Array.unzip
|
|
|> fromArray;
|
|
};
|
|
|
|
let accumulateYs = _transverseShape((aCurrent, aLast) => aCurrent +. aLast);
|
|
let subtractYs = _transverseShape((aCurrent, aLast) => aCurrent -. aLast);
|
|
|
|
module Range = {
|
|
// ((lastX, lastY), (nextX, nextY))
|
|
type zippedRange = ((float, float), (float, float));
|
|
|
|
let floatSum = Belt.Array.reduce(_, 0., (a, b) => a +. b);
|
|
let toT = r => r |> Belt.Array.unzip |> fromArray;
|
|
let nextX = ((_, (nextX, _)): zippedRange) => nextX;
|
|
|
|
let rangeAreaAssumingSteps =
|
|
(((lastX, lastY), (nextX, _)): zippedRange) =>
|
|
(nextX -. lastX) *. lastY;
|
|
|
|
let rangeAreaAssumingTriangles =
|
|
(((lastX, lastY), (nextX, nextY)): zippedRange) =>
|
|
(nextX -. lastX) *. (lastY +. nextY) /. 2.;
|
|
|
|
let delta_y_over_delta_x =
|
|
(((lastX, lastY), (nextX, nextY)): zippedRange) =>
|
|
(nextY -. lastY) /. (nextX -. lastX);
|
|
|
|
let inRanges = (mapper, reducer, t: t) => {
|
|
Belt.Array.zip(t.xs, t.ys)
|
|
|> E.A.toRanges
|
|
|> E.R.toOption
|
|
|> E.O.fmap(r => r |> Belt.Array.map(_, mapper) |> reducer);
|
|
};
|
|
|
|
let mapYsBasedOnRanges = fn => inRanges(r => (nextX(r), fn(r)), toT);
|
|
|
|
let integrateWithSteps = z =>
|
|
mapYsBasedOnRanges(rangeAreaAssumingSteps, z) |> E.O.fmap(accumulateYs);
|
|
|
|
let integrateWithTriangles = z =>
|
|
mapYsBasedOnRanges(rangeAreaAssumingTriangles, z)
|
|
|> E.O.fmap(accumulateYs);
|
|
|
|
let derivative = mapYsBasedOnRanges(delta_y_over_delta_x);
|
|
};
|
|
|
|
let findY = CdfLibrary.Distribution.findY;
|
|
let findX = CdfLibrary.Distribution.findX;
|
|
};
|
|
|
|
module Continuous = {
|
|
let minX = XYShape.minX;
|
|
let maxX = XYShape.maxX;
|
|
let fromArrays = XYShape.fromArrays;
|
|
let toJs = XYShape.toJs;
|
|
let toPdf = XYShape.Range.derivative;
|
|
let toCdf = XYShape.Range.integrateWithTriangles;
|
|
let findX = CdfLibrary.Distribution.findX;
|
|
let findY = CdfLibrary.Distribution.findY;
|
|
let findIntegralY = (f, r) => {
|
|
r |> toCdf |> E.O.fmap(findY(f));
|
|
};
|
|
|
|
let normalizeCdf = (continuousShape: continuousShape) =>
|
|
continuousShape |> XYShape.scaleCdfTo(~scaleTo=1.0);
|
|
|
|
let scalePdf = (~scaleTo=1.0, continuousShape: continuousShape) =>
|
|
continuousShape
|
|
|> toCdf
|
|
|> E.O.fmap(XYShape.scaleCdfTo(~scaleTo))
|
|
|> E.O.bind(_, toPdf);
|
|
};
|
|
|
|
module Discrete = {
|
|
let minX = XYShape.minX;
|
|
let maxX = XYShape.maxX;
|
|
type t = discreteShape;
|
|
let fromArrays = XYShape.fromArrays;
|
|
let toJs = XYShape.toJs;
|
|
let ySum = XYShape.ySum;
|
|
let zip = t => Belt.Array.zip(t.xs, t.ys);
|
|
|
|
let scaleYToTotal = (totalDesired, t: t): t => {
|
|
let difference = totalDesired /. ySum(t);
|
|
XYShape.fmap(t, y => y *. difference);
|
|
};
|
|
|
|
let render = (t: t) =>
|
|
Belt.Array.zip(t.xs, t.ys)
|
|
|> E.A.fmap(((x, y)) =>
|
|
<div>
|
|
{E.Float.toFixed(x)
|
|
++ "---"
|
|
++ E.Float.with3DigitsPrecision(y *. 100.)
|
|
|> ReasonReact.string}
|
|
</div>
|
|
)
|
|
|> ReasonReact.array;
|
|
|
|
let integrate = XYShape.accumulateYs;
|
|
let derivative = XYShape.subtractYs;
|
|
|
|
// TODO: This has a clear bug where it returns the Y value of the first item,
|
|
// even if X is less than the X of the first item.
|
|
// It has a second bug that it assumes things are triangular, instead of interpolating via steps.
|
|
let findIntegralY = (f, t: t) => {
|
|
t |> XYShape.accumulateYs |> CdfLibrary.Distribution.findY(f);
|
|
};
|
|
|
|
let findY = (f, t: t) => {
|
|
Belt.Array.zip(t.xs, t.ys)
|
|
|> E.A.getBy(_, ((x, _)) => x == f)
|
|
|> E.O.fmap(((_, y)) => y)
|
|
|> E.O.default(0.);
|
|
};
|
|
};
|
|
|
|
let min = (f1: option(float), f2: option(float)) =>
|
|
switch (f1, f2) {
|
|
| (Some(f1), Some(f2)) => Some(f1 < f2 ? f1 : f2)
|
|
| (Some(f1), None) => Some(f1)
|
|
| (None, Some(f2)) => Some(f2)
|
|
| (None, None) => None
|
|
};
|
|
|
|
let max = (f1: option(float), f2: option(float)) =>
|
|
switch (f1, f2) {
|
|
| (Some(f1), Some(f2)) => Some(f1 > f2 ? f1 : f2)
|
|
| (Some(f1), None) => Some(f1)
|
|
| (None, Some(f2)) => Some(f2)
|
|
| (None, None) => None
|
|
};
|
|
|
|
module Mixed = {
|
|
let make = (~continuous, ~discrete, ~discreteProbabilityMassFraction) => {
|
|
continuous,
|
|
discrete,
|
|
discreteProbabilityMassFraction,
|
|
};
|
|
|
|
let minX = (t: DistributionTypes.mixedShape) =>
|
|
min(t.continuous |> Continuous.minX, t.discrete |> Discrete.minX);
|
|
|
|
let maxX = (t: DistributionTypes.mixedShape) => {
|
|
max(t.continuous |> Continuous.maxX, t.discrete |> Discrete.maxX);
|
|
};
|
|
|
|
let mixedMultiply =
|
|
(
|
|
t: DistributionTypes.mixedShape,
|
|
continuousComponent,
|
|
discreteComponent,
|
|
) => {
|
|
let diffFn = t.discreteProbabilityMassFraction;
|
|
continuousComponent *. (1.0 -. diffFn) +. discreteComponent *. diffFn;
|
|
};
|
|
|
|
type yPdfPoint = {
|
|
continuous: option(float),
|
|
discrete: option(float),
|
|
discreteProbabilityMassFraction: float,
|
|
};
|
|
|
|
let findY = (t: DistributionTypes.mixedShape, x: float): yPdfPoint => {
|
|
continuous: Continuous.findY(x, t.continuous) |> E.O.some,
|
|
discrete: Discrete.findY(x, t.discrete) |> E.O.some,
|
|
discreteProbabilityMassFraction: t.discreteProbabilityMassFraction,
|
|
};
|
|
|
|
let findYIntegral =
|
|
(x: float, t: DistributionTypes.mixedShape): option(float) => {
|
|
let c = t.continuous |> Continuous.findIntegralY(x);
|
|
let d = Discrete.findIntegralY(x, t.discrete);
|
|
switch (c, d) {
|
|
| (Some(c), d) => Some(mixedMultiply(t, c, d))
|
|
| _ => None
|
|
};
|
|
};
|
|
};
|
|
|
|
module Any = {
|
|
type t = DistributionTypes.pointsType;
|
|
|
|
let y = (t: t, x: float) =>
|
|
switch (t) {
|
|
| Mixed(m) => `mixed(Mixed.findY(m, x))
|
|
| Discrete(discreteShape) => `discrete(Discrete.findY(x, discreteShape))
|
|
| Continuous(continuousShape) =>
|
|
`continuous(Continuous.findY(x, continuousShape))
|
|
};
|
|
|
|
let yIntegral = (t: t, x: float) =>
|
|
switch (t) {
|
|
| Mixed(m) => Mixed.findYIntegral(x, m)
|
|
| Discrete(discreteShape) =>
|
|
Discrete.findIntegralY(x, discreteShape) |> E.O.some
|
|
| Continuous(continuousShape) =>
|
|
Continuous.findIntegralY(x, continuousShape)
|
|
};
|
|
|
|
let minX = (t: t) =>
|
|
switch (t) {
|
|
| Mixed(m) => Mixed.minX(m)
|
|
| Discrete(discreteShape) => Discrete.minX(discreteShape)
|
|
| Continuous(continuousShape) => Continuous.minX(continuousShape)
|
|
};
|
|
|
|
let maxX = (t: t) =>
|
|
switch (t) {
|
|
| Mixed(m) => Mixed.maxX(m)
|
|
| Discrete(discreteShape) => Discrete.maxX(discreteShape)
|
|
| Continuous(continuousShape) => Continuous.maxX(continuousShape)
|
|
};
|
|
|
|
// TODO: This is wrong. The discrete component should be made continuous when integrating.
|
|
let pdfToCdf = (t: t) =>
|
|
switch (t) {
|
|
| Mixed({continuous, discrete, discreteProbabilityMassFraction}) =>
|
|
Some(
|
|
Mixed({
|
|
continuous: Continuous.toCdf(continuous) |> E.O.toExt(""),
|
|
discrete: discrete |> Discrete.integrate,
|
|
discreteProbabilityMassFraction,
|
|
}),
|
|
)
|
|
| Discrete(discrete) => Some(Continuous(discrete |> Discrete.integrate))
|
|
| Continuous(continuous) =>
|
|
Continuous.toCdf(continuous) |> E.O.fmap(e => Continuous(e))
|
|
};
|
|
};
|
|
|
|
module DomainMixed = {
|
|
type t = {
|
|
mixedShape,
|
|
domain,
|
|
};
|
|
}; |