[api]: Improve Curve Fitting algorithm

tiagohm · May 22, 2024 · 617bf57 · 617bf57
1 parent 29502bc
commit 617bf57
Show file tree

Hide file tree

Showing 9 changed files with 26 additions and 30 deletions.
diff --git a/nebulosa-curve-fitting/src/main/kotlin/nebulosa/curve/fitting/Curve.kt b/nebulosa-curve-fitting/src/main/kotlin/nebulosa/curve/fitting/Curve.kt
@@ -1,15 +1,15 @@
 package nebulosa.curve.fitting
 
 import org.apache.commons.math3.analysis.UnivariateFunction
-import org.apache.commons.math3.analysis.polynomials.PolynomialFunction
 
 fun interface Curve : UnivariateFunction {
 
     operator fun invoke(x: Double) = value(x)
 
     companion object {
 
-        internal fun DoubleArray.curvePoints(): Collection<CurvePoint> {
+        @JvmStatic
+        fun DoubleArray.curvePoints(): Collection<CurvePoint> {
             val points = ArrayList<CurvePoint>(size / 2)
 
             for (i in indices step 2) {
@@ -18,8 +18,5 @@ fun interface Curve : UnivariateFunction {
 
             return points
         }
-
-        @Suppress("NOTHING_TO_INLINE")
-        internal inline fun DoubleArray.polynomial() = PolynomialFunction(this)
     }
 }
diff --git a/nebulosa-curve-fitting/src/main/kotlin/nebulosa/curve/fitting/FittedCurve.kt b/nebulosa-curve-fitting/src/main/kotlin/nebulosa/curve/fitting/FittedCurve.kt
@@ -5,10 +5,4 @@ interface FittedCurve : Curve {
     val minimum: CurvePoint
 
     val rSquared: Double
-
-    val minimumX
-        get() = minimum.x
-
-    val minimumY
-        get() = minimum.y
 }
diff --git a/nebulosa-curve-fitting/src/main/kotlin/nebulosa/curve/fitting/PolynomialCurve.kt b/nebulosa-curve-fitting/src/main/kotlin/nebulosa/curve/fitting/PolynomialCurve.kt
@@ -0,0 +1,10 @@
+package nebulosa.curve.fitting
+
+import org.apache.commons.math3.analysis.UnivariateFunction
+
+interface PolynomialCurve : Curve {
+
+    val polynomial: UnivariateFunction
+
+    override fun value(x: Double) = polynomial.value(x)
+}
diff --git a/nebulosa-curve-fitting/src/main/kotlin/nebulosa/curve/fitting/QuadraticFitting.kt b/nebulosa-curve-fitting/src/main/kotlin/nebulosa/curve/fitting/QuadraticFitting.kt
@@ -1,21 +1,18 @@
 package nebulosa.curve.fitting
 
-import nebulosa.curve.fitting.Curve.Companion.polynomial
 import org.apache.commons.math3.analysis.UnivariateFunction
+import org.apache.commons.math3.analysis.polynomials.PolynomialFunction
 import org.apache.commons.math3.fitting.PolynomialCurveFitter
 
 data object QuadraticFitting : CurveFitting<QuadraticFitting.Curve> {
 
     data class Curve(
-        private val poly: UnivariateFunction,
+        override val polynomial: UnivariateFunction,
         override val minimum: CurvePoint,
         override val rSquared: Double,
-    ) : FittedCurve {
+    ) : FittedCurve, PolynomialCurve
 
-        override fun value(x: Double) = poly.value(x)
-    }
-
-    override fun calculate(points: Collection<CurvePoint>) = with(FITTER.fit(points).polynomial()) {
+    override fun calculate(points: Collection<CurvePoint>) = with(PolynomialFunction(FITTER.fit(points))) {
         val rSquared = RSquared.calculate(points, this)
         val minimumX = coefficients[1] / (-2.0 * coefficients[2])
         val minimumY = value(minimumX)

diff --git a/nebulosa-curve-fitting/src/main/kotlin/nebulosa/curve/fitting/RSquared.kt b/nebulosa-curve-fitting/src/main/kotlin/nebulosa/curve/fitting/RSquared.kt
@@ -6,6 +6,7 @@ import kotlin.math.pow
 
 object RSquared {
 
+    @JvmStatic
     fun calculate(points: Collection<CurvePoint>, function: UnivariateFunction): Double {
         val descriptiveStatistics = DescriptiveStatistics(points.size)
         val predictedValues = DoubleArray(points.size)

diff --git a/nebulosa-curve-fitting/src/main/kotlin/nebulosa/curve/fitting/TrendLine.kt b/nebulosa-curve-fitting/src/main/kotlin/nebulosa/curve/fitting/TrendLine.kt
@@ -13,19 +13,16 @@ data class TrendLine(val points: Collection<CurvePoint>) : LinearCurve {
         points.forEach { regression.addData(it.x, it.y) }
     }
 
-    override val slope = regression.slope.zeroIfNaN()
+    override val slope = regression.slope.let { if (it.isNaN()) 0.0 else it }
 
-    override val intercept = regression.intercept.zeroIfNaN()
+    override val intercept = regression.intercept.let { if (it.isNaN()) 0.0 else it }
 
-    override val rSquared = regression.rSquare.zeroIfNaN()
+    override val rSquared = regression.rSquare.let { if (it.isNaN()) 0.0 else it }
 
-    override fun value(x: Double) = regression.predict(x)
+    override fun value(x: Double) = if (points.isEmpty()) 0.0 else regression.predict(x)
 
     companion object {
 
         @JvmStatic val ZERO = TrendLine()
-
-        @Suppress("NOTHING_TO_INLINE")
-        private inline fun Double.zeroIfNaN() = if (isNaN()) 0.0 else this
     }
 }
diff --git a/nebulosa-curve-fitting/src/test/kotlin/HyperbolicFittingTest.kt b/nebulosa-curve-fitting/src/test/kotlin/HyperbolicFittingTest.kt
@@ -14,8 +14,8 @@ class HyperbolicFittingTest : StringSpec(), CurveFitting<HyperbolicFitting.Curve
                 6.0, 3.0, 7.0, 6.0, 8.0, 11.0, 9.0, 18.0,
             )
 
-            curve.minimumX shouldBe (5.0 plusOrMinus 1e-12)
-            curve.minimumY shouldBe (1.2 plusOrMinus 1e-12)
+            curve.minimum.x shouldBe (5.0 plusOrMinus 1e-12)
+            curve.minimum.y shouldBe (1.2 plusOrMinus 1e-12)
         }
         "bad data:prevent infinit loop" {
             shouldThrow<IllegalArgumentException> { calculate(1000.0, 18.0, 1100.0, 0.0, 1200.0, 0.0) }

diff --git a/nebulosa-curve-fitting/src/test/kotlin/QuadraticFittingTest.kt b/nebulosa-curve-fitting/src/test/kotlin/QuadraticFittingTest.kt
@@ -17,8 +17,8 @@ class QuadraticFittingTest : StringSpec(), CurveFitting<QuadraticFitting.Curve>
             )
 
             curve(5.0) shouldBeExactly 2.0
-            curve.minimumX shouldBe (5.0 plusOrMinus 1e-12)
-            curve.minimumY shouldBe (2.0 plusOrMinus 1e-12)
+            curve.minimum.x shouldBe (5.0 plusOrMinus 1e-12)
+            curve.minimum.y shouldBe (2.0 plusOrMinus 1e-12)
             curve.rSquared shouldBe (1.0 plusOrMinus 1e-12)
         }
     }

diff --git a/nebulosa-curve-fitting/src/test/kotlin/TrendLineTest.kt b/nebulosa-curve-fitting/src/test/kotlin/TrendLineTest.kt
@@ -6,7 +6,7 @@ class TrendLineTest : StringSpec() {
 
     init {
         "no points" {
-            val line = TrendLine()
+            val line = TrendLine.ZERO
 
             line.slope shouldBeExactly 0.0
             line.intercept shouldBeExactly 0.0