uncertain: Manipulating numbers with inherent experimental/measurement uncertainty

[ bsd3, library, math ] [ Propose Tags ]

See README.md.

Documentation maintained at https://mstksg.github.io/uncertain


[Skip to Readme]

Downloads

Maintainer's Corner

Package maintainers

For package maintainers and hackage trustees

Candidates

Versions [RSS] 0.2.0.0, 0.3.0.0, 0.3.1.0, 0.4.0.0, 0.4.0.1
Change log CHANGELOG.md
Dependencies ad (>=4), base (>=4.6 && <5), base-compat, containers (>=0.5), free (>=4), mwc-random (>=0.10), primitive (>=0.1), transformers (>=0.2) [details]
License BSD-3-Clause
Copyright (c) Justin Le 2016
Author Justin Le
Maintainer justin@jle.im
Category Math
Home page https://github.com/mstksg/uncertain
Bug tracker https://github.com/mstksg/uncertain/issues
Source repo head: git clone git://github.com/mstksg/uncertain.git
Uploaded by jle at 2016-05-22T09:17:42Z
Distributions NixOS:0.3.1.0, Stackage:0.4.0.1
Reverse Dependencies 1 direct, 0 indirect [details]
Downloads 2613 total (23 in the last 30 days)
Rating (no votes yet) [estimated by Bayesian average]
Your Rating
  • λ
  • λ
  • λ
Status Docs available [build log]
Last success reported on 2016-11-25 [all 1 reports]

Readme for uncertain-0.3.1.0

[back to package description]

Uncertain

Build Status

Provides tools to manipulate numbers with inherent experimental/measurement uncertainty, and propagates them through functions based on principles from statistics.

Documentation maintained at https://mstksg.github.io/uncertain.

Usage

import Numeric.Uncertain

Create numbers

7.13 +/- 0.05
91800 +/- 100
12.5 `withVar` 0.36
exact 7.9512
81.42 `withPrecision` 4
7    :: Uncert Double
9.18 :: Uncert Double
fromSamples [12.5, 12.7, 12.6, 12.6, 12.5]

Can be descontructed/analyzed with :+/- (pattern synonym/pseudo-constructor matching on the mean and standard deviation), uMean, uStd, uVar, etc.

Manipulate with error propagation

ghci> let x = 1.52 +/- 0.07
ghci> let y = 781.4 +/- 0.3
ghci> let z = 1.53e-1 `withPrecision` 3
ghci> cosh x
2.4 +/- 0.2
ghci> exp x / z * sin (y ** z)
10.9 +/- 0.9
ghci> pi + 3 * logBase x y
52 +/- 5

Propagates uncertainty using second-order multivariate Taylor expansions of functions, computed using the ad library.

Arbitrary numeric functions

ghci> liftUF (\[x,y,z] -> x*y+z)
             [ 12.2 +/- 0.5
             , 56 +/- 2
             , 0.12 +/- 0.08
             ]
680 +/- 40

Correlated samples

Can propagate uncertainty on complex functions take from potentially correlated samples.

ghci> import Numeric.Uncertain.Correlated
ghci> evalCorr $ do
        x <- sampleUncert $ 12.5 +/- 0.8
        y <- sampleUncert $ 15.9 +/- 0.5
        z <- sampleUncert $ 1.52 +/- 0.07
        let k = y ** x
        resolveUncert $ (x+z) * logBase z k
1200 +/- 200

"Interactive" Exploratory Mode

Correlated module functionality can be used in ghci or IO or ST, for "interactive" exploration.

ghci> x <- sampleUncert $ 12.5 +/- 0.8
ghci> y <- sampleUncert $ 15.9 +/- 0.5
ghci> z <- sampleUncert $ 1.52 +/- 0.07
ghci> let k = y**x
ghci> resolveUncert $ (x+z) * logBase z k
1200 +/- 200

Monte Carlo-based propagation of uncertainty

Provides a module for propagating uncertainty using Monte Carlo simulations

ghci> import qualified Numeric.Uncertain.MonteCarlo as MC
ghci> import System.Random.MWC
ghci> let x = 1.52 +/- 0.07
ghci> let y = 781.4 +/- 0.3
ghci> let z = 1.53e-1 `withPrecision` 3
ghci> g <- create
ghci> cosh x
2.4 +/- 0.2
ghci> MC.liftU cosh x g
2.4 +/- 0.2
ghci> exp x / z * sin (y ** z)
10.9 +/- 0.9
ghci> MC.liftU3 (\a b c -> exp a / c * sin (b**c)) x y z g
10.8 +/- 1.0
ghci> pi + 3 * logBase x y
52 +/- 5
ghci> MC.liftU2 (\a b -> pi + 3 * logBase a b) x y g
51 +/- 5

Comparisons

Note that this is very different from other libraries with similar data types (like from intervals and rounding); these do not attempt to maintain intervals or simply digit precisions; they instead are intended to model actual experimental and measurement data with their uncertainties, and apply functions to the data with the uncertainties and properly propagating the errors with sound statistical principles.

For a clear example, take

> (52 +/- 6) + (39 +/- 4)
91. +/- 7.

In a library like intervals, this would result in 91 +/- 10 (that is, a lower bound of 46 + 35 and an upper bound of 58 + 43). However, with experimental data, errors in two independent samples tend to "cancel out", and result in an overall aggregate uncertainty in the sum of approximately 7.

Copyright (c) Justin Le 2016