Fixed QOW optimization

examples/src/main/kotlin/space/kscience/kmath/fit/qowFit.kt

@@ -14,6 +14,7 @@ import space.kscience.kmath.expressions.Symbol
 import space.kscience.kmath.expressions.binding
 import space.kscience.kmath.expressions.symbol
 import space.kscience.kmath.optimization.QowOptimizer
+import space.kscience.kmath.optimization.chiSquaredOrNull
 import space.kscience.kmath.optimization.fitWith
 import space.kscience.kmath.optimization.resultPoint
 import space.kscience.kmath.real.map
@@ -50,7 +51,7 @@ suspend fun main() {
 
     //Perform an operation on each x value (much more effective, than numpy)
     val y = x.map { it ->
-        val value = it.pow(2) + it + 100
+        val value = it.pow(2) + it + 1
         value + chain.next() * sqrt(value)
     }
     // this will also work, but less effective:
@@ -63,7 +64,7 @@ suspend fun main() {
     val result = XYErrorColumnarData.of(x, y, yErr).fitWith(
         QowOptimizer,
         DSProcessor,
-        mapOf(a to 1.0, b to 1.2, c to 99.0)
+        mapOf(a to 0.9, b to 1.2, c to 2.0)
     ) { arg ->
         //bind variables to autodiff context
         val a by binding
@@ -96,9 +97,9 @@ suspend fun main() {
         h3 {
             +"Fit result: ${result.resultPoint}"
         }
-//        h3 {
-//            +"Chi2/dof = ${result.resultValue / (x.size - 3)}"
-//        }
+        h3 {
+            +"Chi2/dof = ${result.chiSquaredOrNull!! / (x.size - 3)}"
+        }
     }
 
     page.makeFile()

kmath-optimization/src/commonMain/kotlin/space/kscience/kmath/optimization/QowOptimizer.kt

@@ -50,7 +50,7 @@ public object QowOptimizer : Optimizer<Double, XYFit> {
          */
         val dispersion: Point<Double> by lazy {
             DoubleBuffer(problem.data.size) { d ->
-                problem.weight(d).invoke(parameters)
+                1.0/problem.weight(d).invoke(parameters)
             }
         }
 

kmath-optimization/src/commonMain/kotlin/space/kscience/kmath/optimization/XYFit.kt

@@ -7,6 +7,7 @@
 package space.kscience.kmath.optimization
 
 import space.kscience.kmath.data.XYColumnarData
+import space.kscience.kmath.data.indices
 import space.kscience.kmath.expressions.*
 import space.kscience.kmath.misc.FeatureSet
 import space.kscience.kmath.misc.Loggable
@@ -81,6 +82,7 @@ public class XYFit(
     override val features: FeatureSet<OptimizationFeature>,
     internal val pointToCurveDistance: PointToCurveDistance = PointToCurveDistance.byY,
     internal val pointWeight: PointWeight = PointWeight.byYSigma,
+    public val xSymbol: Symbol = Symbol.x,
 ) : OptimizationProblem<Double> {
     public fun distance(index: Int): DifferentiableExpression<Double> = pointToCurveDistance.distance(this, index)
 
@@ -119,7 +121,26 @@ public suspend fun <I : Any, A> XYColumnarData<Double, Double, Double>.fitWith(
         modelExpression,
         actualFeatures,
         pointToCurveDistance,
-        pointWeight
+        pointWeight,
+        xSymbol
     )
     return optimizer.optimize(problem)
 }
+
+/**
+ * Compute chi squared value for completed fit. Return null for incomplete fit
+ */
+public val XYFit.chiSquaredOrNull: Double? get() {
+    val result = resultPointOrNull ?: return null
+
+    return data.indices.sumOf { index->
+
+        val x = data.x[index]
+        val y = data.y[index]
+        val yErr = data[Symbol.yError]?.get(index) ?: 1.0
+
+        val mu = model.invoke(result + (xSymbol to x) )
+
+        ((y - mu)/yErr).pow(2)
+    }
+}

kmath-optimization/src/commonMain/kotlin/space/kscience/kmath/optimization/logLikelihood.kt

@@ -26,7 +26,7 @@ internal fun XYFit.logLikelihood(): DifferentiableExpression<Double> = object :
         data.indices.sumOf { index ->
             val d = distance(index)(arguments)
             val weight = weight(index)(arguments)
-            val weightDerivative = weight(index)(arguments)
+            val weightDerivative = weight(index).derivative(symbols)(arguments)
 
             //  -1 / (sqrt(2 PI) * sigma) + 2 (x-mu)/ 2 sigma^2 * d mu/ d theta - (x-mu)^2 / 2 * d w/ d theta
             return@sumOf -oneOver2Pi * sqrt(weight) + //offset derivative