1
0
mirror of https://github.com/ppy/osu.git synced 2026-05-17 20:35:39 +08:00
Files
osu-lazer/osu.Game/Rulesets/Difficulty/Utils/DifficultyCalculationUtils.cs
T
Natelytle d5ef8c8524 Replace error functions in DifficultyCalculationUtils with good-enough approximations (#33717)
* Reimplement error functions

* Fix bug with adjustment for negative values

* Formatting

---------

Co-authored-by: tsunyoku <tsunyoku@gmail.com>
2025-06-18 13:14:01 +00:00

195 lines
8.5 KiB
C#

// Copyright (c) ppy Pty Ltd <contact@ppy.sh>. Licensed under the MIT Licence.
// See the LICENCE file in the repository root for full licence text.
using System;
using System.Linq;
namespace osu.Game.Rulesets.Difficulty.Utils
{
public static partial class DifficultyCalculationUtils
{
/// <summary>
/// Converts BPM value into milliseconds
/// </summary>
/// <param name="bpm">Beats per minute</param>
/// <param name="delimiter">Which rhythm delimiter to use, default is 1/4</param>
/// <returns>BPM conveted to milliseconds</returns>
public static double BPMToMilliseconds(double bpm, int delimiter = 4)
{
return 60000.0 / delimiter / bpm;
}
/// <summary>
/// Converts milliseconds value into a BPM value
/// </summary>
/// <param name="ms">Milliseconds</param>
/// <param name="delimiter">Which rhythm delimiter to use, default is 1/4</param>
/// <returns>Milliseconds conveted to beats per minute</returns>
public static double MillisecondsToBPM(double ms, int delimiter = 4)
{
return 60000.0 / (ms * delimiter);
}
/// <summary>
/// Calculates a S-shaped logistic function (https://en.wikipedia.org/wiki/Logistic_function)
/// </summary>
/// <param name="x">Value to calculate the function for</param>
/// <param name="maxValue">Maximum value returnable by the function</param>
/// <param name="multiplier">Growth rate of the function</param>
/// <param name="midpointOffset">How much the function midpoint is offset from zero <paramref name="x"/></param>
/// <returns>The output of logistic function of <paramref name="x"/></returns>
public static double Logistic(double x, double midpointOffset, double multiplier, double maxValue = 1) => maxValue / (1 + Math.Exp(multiplier * (midpointOffset - x)));
/// <summary>
/// Calculates a S-shaped logistic function (https://en.wikipedia.org/wiki/Logistic_function)
/// </summary>
/// <param name="maxValue">Maximum value returnable by the function</param>
/// <param name="exponent">Exponent</param>
/// <returns>The output of logistic function</returns>
public static double Logistic(double exponent, double maxValue = 1) => maxValue / (1 + Math.Exp(exponent));
/// <summary>
/// Returns the <i>p</i>-norm of an <i>n</i>-dimensional vector (https://en.wikipedia.org/wiki/Norm_(mathematics))
/// </summary>
/// <param name="p">The value of <i>p</i> to calculate the norm for.</param>
/// <param name="values">The coefficients of the vector.</param>
/// <returns>The <i>p</i>-norm of the vector.</returns>
public static double Norm(double p, params double[] values) => Math.Pow(values.Sum(x => Math.Pow(x, p)), 1 / p);
/// <summary>
/// Calculates a Gaussian-based bell curve function (https://en.wikipedia.org/wiki/Gaussian_function)
/// </summary>
/// <param name="x">Value to calculate the function for</param>
/// <param name="mean">The mean (center) of the bell curve</param>
/// <param name="width">The width (spread) of the curve</param>
/// <param name="multiplier">Multiplier to adjust the curve's height</param>
/// <returns>The output of the bell curve function of <paramref name="x"/></returns>
public static double BellCurve(double x, double mean, double width, double multiplier = 1.0) => multiplier * Math.Exp(Math.E * -(Math.Pow(x - mean, 2) / Math.Pow(width, 2)));
/// <summary>
/// Calculates a Smoothstep Bellcurve that returns returns 1 for x = mean, and smoothly reducing it's value to 0 over width
/// </summary>
/// <param name="x">Value to calculate the function for</param>
/// <param name="mean">Value of x, for which return value will be the highest (=1)</param>
/// <param name="width">Range [mean - width, mean + width] where function will change values</param>
/// <returns>The output of the smoothstep bell curve function of <paramref name="x"/></returns>
public static double SmoothstepBellCurve(double x, double mean = 0.5, double width = 0.5)
{
x -= mean;
x = x > 0 ? (width - x) : (width + x);
return Smoothstep(x, 0, width);
}
/// <summary>
/// Smoothstep function (https://en.wikipedia.org/wiki/Smoothstep)
/// </summary>
/// <param name="x">Value to calculate the function for</param>
/// <param name="start">Value at which function returns 0</param>
/// <param name="end">Value at which function returns 1</param>
public static double Smoothstep(double x, double start, double end)
{
x = Math.Clamp((x - start) / (end - start), 0.0, 1.0);
return x * x * (3.0 - 2.0 * x);
}
/// <summary>
/// Smootherstep function (https://en.wikipedia.org/wiki/Smoothstep#Variations)
/// </summary>
/// <param name="x">Value to calculate the function for</param>
/// <param name="start">Value at which function returns 0</param>
/// <param name="end">Value at which function returns 1</param>
public static double Smootherstep(double x, double start, double end)
{
x = Math.Clamp((x - start) / (end - start), 0.0, 1.0);
return x * x * x * (x * (6.0 * x - 15.0) + 10.0);
}
/// <summary>
/// Reverse linear interpolation function (https://en.wikipedia.org/wiki/Linear_interpolation)
/// </summary>
/// <param name="x">Value to calculate the function for</param>
/// <param name="start">Value at which function returns 0</param>
/// <param name="end">Value at which function returns 1</param>
public static double ReverseLerp(double x, double start, double end)
{
return Math.Clamp((x - start) / (end - start), 0.0, 1.0);
}
/// <summary>
/// Error function (https://en.wikipedia.org/wiki/Error_function)
/// </summary>
/// <param name="x">Value to calculate the function for</param>
public static double Erf(double x)
{
if (x == 0)
return 0;
if (double.IsPositiveInfinity(x))
return 1;
if (double.IsNegativeInfinity(x))
return -1;
if (double.IsNaN(x))
return double.NaN;
// Constants for approximation (Abramowitz and Stegun formula 7.1.26)
double t = 1.0 / (1.0 + 0.3275911 * Math.Abs(x));
double tau = t * (0.254829592
+ t * (-0.284496736
+ t * (1.421413741
+ t * (-1.453152027
+ t * 1.061405429))));
double erf = 1.0 - tau * Math.Exp(-x * x);
return x >= 0 ? erf : -erf;
}
/// <summary>
/// Complementary error function (https://en.wikipedia.org/wiki/Error_function)
/// </summary>
/// <param name="x">Value to calculate the function for</param>
public static double Erfc(double x) => 1 - Erf(x);
/// <summary>
/// Inverse error function (https://en.wikipedia.org/wiki/Error_function)
/// </summary>
/// <param name="x">Value to calculate the function for</param>
public static double ErfInv(double x)
{
if (x <= -1)
return double.NegativeInfinity;
if (x >= 1)
return double.PositiveInfinity;
if (x == 0)
return 0;
const double a = 0.147;
double sgn = Math.Sign(x);
x = Math.Abs(x);
double ln = Math.Log(1 - x * x);
double t1 = 2 / (Math.PI * a) + ln / 2;
double t2 = ln / a;
double baseApprox = Math.Sqrt(t1 * t1 - t2) - t1;
// Correction reduces max error from -0.005 to -0.00045.
double c = x >= 0.85 ? Math.Pow((x - 0.85) / 0.293, 8) : 0;
double erfInv = sgn * (Math.Sqrt(baseApprox) + c);
return erfInv;
}
/// <summary>
/// Inverse complementary error function (https://en.wikipedia.org/wiki/Error_function)
/// </summary>
/// <param name="x">Value to calculate the function for</param>
public static double ErfcInv(double x) => ErfInv(1 - x);
}
}