quantified.uncertainty/squiggle
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Fix multiplication of variances in ShapeConvolution

Browse Source
This commit is contained in:
Sebastian Kosch 2020-07-03 17:13:26 -07:00
parent ca9f725ae7
commit 730dbddaf9
1 changed files with 23 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

33
src/distPlus/distribution/ShapeConvolution.re
View File

@ -80,15 +80,20 @@ let toDiscretePointMassesFromTriangulars =
{n: n - 2, masses, means, variances}; {n: n - 2, masses, means, variances};
} else { } else {
for (i in 1 to n - 2) { for (i in 1 to n - 2) {
// area of triangle = width * height / 2
let _ = let _ =
Belt.Array.set( Belt.Array.set(
masses, masses,
i - 1, i - 1,
(xs[i + 1] -. xs[i - 1]) *. ys[i] /. 2., (xs[i + 1] -. xs[i - 1]) *. ys[i] /. 2.,
); );
// means of triangle = (a + b + c) / 3
let _ = let _ =
Belt.Array.set(means, i - 1, (xs[i - 1] +. xs[i] +. xs[i + 1]) /. 3.); Belt.Array.set(means, i - 1, (xs[i - 1] +. xs[i] +. xs[i + 1]) /. 3.);
// variance of triangle = (a^2 + b^2 + c^2 - ab - ac - bc) / 18
let _ = let _ =
Belt.Array.set( Belt.Array.set(
variances, variances,
@ -118,7 +123,10 @@ let combineShapesContinuousContinuous =
// if we add the two distributions, we should probably use normal filters. // if we add the two distributions, we should probably use normal filters.
// if we multiply the two distributions, we should probably use lognormal filters. // if we multiply the two distributions, we should probably use lognormal filters.
let t1m = toDiscretePointMassesFromTriangulars(s1); let t1m = toDiscretePointMassesFromTriangulars(s1);
let t2m = toDiscretePointMassesFromTriangulars(s2); let t2m = switch (op) {
| `Divide => toDiscretePointMassesFromTriangulars(~inverse=true, s2)
| _ => toDiscretePointMassesFromTriangulars(~inverse=false, s2)
};
let combineMeansFn = let combineMeansFn =
switch (op) { switch (op) {
@ -134,7 +142,7 @@ let combineShapesContinuousContinuous =
| `Add => ((v1, v2, m1, m2) => v1 +. v2) | `Add => ((v1, v2, m1, m2) => v1 +. v2)
| `Subtract => ((v1, v2, m1, m2) => v1 +. v2) | `Subtract => ((v1, v2, m1, m2) => v1 +. v2)
| `Multiply => ( | `Multiply => (
(v1, v2, m1, m2) => v1 *. v2 +. v1 *. m1 ** 2. +. v2 *. m1 ** 2. (v1, v2, m1, m2) => v1 *. v2 +. v1 *. m2 ** 2. +. v2 *. m1 ** 2.
) )
| `Divide => ( | `Divide => (
(v1, vInv2, m1, mInv2) => (v1, vInv2, m1, mInv2) =>
@ -142,6 +150,7 @@ let combineShapesContinuousContinuous =
) )
}; };
// TODO: If operating on two positive-domain distributions, we should take that into account
let outputMinX: ref(float) = ref(infinity); let outputMinX: ref(float) = ref(infinity);
let outputMaxX: ref(float) = ref(neg_infinity); let outputMaxX: ref(float) = ref(neg_infinity);
let masses: array(float) = let masses: array(float) =
@ -180,16 +189,18 @@ let combineShapesContinuousContinuous =
// we now want to create a set of target points. For now, let's just evenly distribute 200 points between // we now want to create a set of target points. For now, let's just evenly distribute 200 points between
// between the outputMinX and outputMaxX // between the outputMinX and outputMaxX
let outputXs: array(float) = let nOut = 300;
E.A.Floats.range(outputMinX^, outputMaxX^, 200); let outputXs: array(float) = E.A.Floats.range(outputMinX^, outputMaxX^, nOut);
let outputYs: array(float) = Belt.Array.make(200, 0.0); let outputYs: array(float) = Belt.Array.make(nOut, 0.0);
// now, for each of the outputYs, accumulate from a Gaussian kernel over each input point. // now, for each of the outputYs, accumulate from a Gaussian kernel over each input point.
for (i in 0 to E.A.length(outputXs) - 1) { for (j in 0 to E.A.length(masses) - 1) {
for (j in 0 to E.A.length(masses) - 1) { let _ = if (variances[j] > 0.) {
let dx = outputXs[i] -. means[j]; for (i in 0 to E.A.length(outputXs) - 1) {
let contribution = let dx = outputXs[i] -. means[j];
masses[j] *. exp(-. (dx ** 2.) /. (2. *. variances[j])); let contribution = masses[j] *. exp(-. (dx ** 2.) /. (2. *. variances[j]));
let _ = Belt.Array.set(outputYs, i, outputYs[i] +. contribution); let _ = Belt.Array.set(outputYs, i, outputYs[i] +. contribution);
();
};
(); ();
}; };
(); ();