Remaking descriptors

This commit is contained in:
Alexander Nozik 2018-07-04 21:43:54 +03:00
parent d8913bb984
commit 1d524f46c8
8 changed files with 510 additions and 22 deletions

View File

@ -28,6 +28,7 @@ allprojects {
//maven { url = "https://jitpack.io" }
maven { url = "http://dl.bintray.com/kotlin/ktor" }
maven { url = "https://dl.bintray.com/kotlin/kotlinx" }
maven { url = "https://dl.bintray.com/kotlin/kotlin-eap"}
}
dependencies {

View File

@ -146,7 +146,7 @@ class PKT8Display : DeviceDisplayFX<PKT8Device>(), PKT8ValueListener {
private val plotFrame: PlotFrame by lazy {
JFreeChartFrame(plotFrameMeta).apply {
plots.descriptor = Descriptors.forElement(TimePlot::class.java)
plots.descriptor = Descriptors.forType(TimePlot::class.java)
PlotUtils.setXAxis(this, "timestamp", "", "time")
}
}

View File

@ -61,8 +61,10 @@ class SpectrumAdapter : BasicAdapter {
}
fun buildSpectrumDataPoint(x: Double, count: Long, countErr: Double, t: Double): Values {
return ValueMap.of(arrayOf(getComponentName(X_VALUE_KEY), getComponentName(Y_VALUE_KEY), getComponentName(Y_ERROR_KEY), getComponentName(POINT_LENGTH_NAME)),
x, count, countErr, t)
return ValueMap.of(
arrayOf(getComponentName(X_VALUE_KEY), getComponentName(Y_VALUE_KEY), getComponentName(Y_ERROR_KEY), getComponentName(POINT_LENGTH_NAME)),
x, count, countErr, t
)
}

View File

@ -0,0 +1,417 @@
/*
* Copyright 2018 Alexander Nozik.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package inr.numass.models.misc
import hep.dataforge.maths.functions.FunctionCaching
import hep.dataforge.maths.integration.GaussRuleIntegrator
import hep.dataforge.plots.PlotFrame
import hep.dataforge.plots.XYFunctionPlot
import hep.dataforge.utils.Misc
import hep.dataforge.values.Values
import org.apache.commons.math3.analysis.BivariateFunction
import org.apache.commons.math3.analysis.UnivariateFunction
import org.apache.commons.math3.exception.OutOfRangeException
import org.slf4j.LoggerFactory
import java.lang.Math.exp
import java.util.*
/**
* Вычисление произвольного порядка функции рассеяния. Не учитывается
* зависимость сечения от энергии электрона
*
* @author Darksnake
*/
object LossCalculator {
private val cache = HashMap<Int, UnivariateFunction>().also {
it[1] = singleScatterFunction
}
fun getGunLossProbabilities(X: Double): List<Double> {
val res = ArrayList<Double>()
var prob: Double
if (X > 0) {
prob = Math.exp(-X)
} else {
// если x ==0, то выживает только нулевой член, первый равен 1
res.add(1.0)
return res
}
res.add(prob)
var n = 0
while (prob > SCATTERING_PROBABILITY_THRESHOLD) {
/*
* prob(n) = prob(n-1)*X/n;
*/
n++
prob *= X / n
res.add(prob)
}
return res
}
fun getGunZeroLossProb(X: Double): Double {
return Math.exp(-X)
}
/**
* Ленивое рекурсивное вычисление функции потерь через предыдущие
*
* @param order
* @return
*/
private fun getLoss(order: Int): UnivariateFunction? {
if (order <= 0) {
throw IllegalArgumentException()
}
return if (cache.containsKey(order)) {
cache[order]
} else {
synchronized(this) {
cache.getOrPut(order) {
LoggerFactory.getLogger(javaClass)
.debug("Scatter cache of order {} not found. Updating", order)
getNextLoss(getMargin(order), getLoss(order - 1))
}
return cache[order]
}
}
}
fun getLossFunction(order: Int): BivariateFunction {
assert(order > 0)
return BivariateFunction { Ei: Double, Ef: Double -> getLossValue(order, Ei, Ef) }
}
fun getLossProbDerivs(x: Double): List<Double> {
val res = ArrayList<Double>()
val probs = getLossProbabilities(x)
var delta = Math.exp(-x)
res.add((delta - probs[0]) / x)
for (i in 1 until probs.size) {
delta *= x / i
res.add((delta - probs[i]) / x)
}
return res
}
/**
* рекурсивно вычисляем все вероятности, котрорые выше порога
*
*
* дисер, стр.48
*
* @param X
* @return
*/
fun getLossProbabilities(x: Double): List<Double> {
return (lossProbCache).getOrPut(x) {
val res = ArrayList<Double>()
var prob: Double
if (x > 0) {
prob = 1 / x * (1 - Math.exp(-x))
} else {
// если x ==0, то выживает только нулевой член, первый равен нулю
res.add(1.0)
return@getOrPut res
}
res.add(prob)
while (prob > SCATTERING_PROBABILITY_THRESHOLD) {
/*
* prob(n) = prob(n-1)-1/n! * X^n * exp(-X);
*/
var delta = Math.exp(-x)
for (i in 1 until res.size + 1) {
delta *= x / i
}
prob -= delta / x
res.add(prob)
}
res
}
}
fun getLossProbability(order: Int, X: Double): Double {
if (order == 0) {
return if (X > 0) {
1 / X * (1 - Math.exp(-X))
} else {
1.0
}
}
val probs = getLossProbabilities(X)
return if (order >= probs.size) {
0.0
} else {
probs[order]
}
}
fun getLossValue(order: Int, Ei: Double, Ef: Double): Double {
return if (Ei - Ef < 5.0) {
0.0
} else if (Ei - Ef >= getMargin(order)) {
0.0
} else {
getLoss(order)!!.value(Ei - Ef)
}
}
/**
* функция потерь с произвольными вероятностями рассеяния
*
* @param probs
* @param Ei
* @param Ef
* @return
*/
fun getLossValue(probs: List<Double>, Ei: Double, Ef: Double): Double {
var sum = 0.0
for (i in 1 until probs.size) {
sum += probs[i] * getLossValue(i, Ei, Ef)
}
return sum
}
/**
* граница интегрирования
*
* @param order
* @return
*/
private fun getMargin(order: Int): Double {
return 80 + order * 50.0
}
/**
* генерирует кэшированную функцию свертки loss со спектром однократных
* потерь
*
* @param loss
* @return
*/
private fun getNextLoss(margin: Double, loss: UnivariateFunction?): UnivariateFunction {
val res = { x: Double ->
val integrand = UnivariateFunction { y: Double ->
try {
loss!!.value(x - y) * singleScatterFunction.value(y)
} catch (ex: OutOfRangeException) {
0.0
}
}
integrator.integrate(5.0, margin, integrand)
}
return FunctionCaching.cacheUnivariateFunction(0.0, margin, 200, res)
}
fun getTotalLossBivariateFunction(X: Double): BivariateFunction {
return BivariateFunction { Ei: Double, Ef: Double -> getTotalLossValue(X, Ei, Ef) }
}
/**
* Значение полной производной функции потерь с учетом всех неисчезающих
* порядков
*
* @param X
* @param Ei
* @param Ef
* @return
*/
fun getTotalLossDeriv(X: Double, Ei: Double, Ef: Double): Double {
val probs = getLossProbDerivs(X)
var sum = 0.0
for (i in 1 until probs.size) {
sum += probs[i] * getLossValue(i, Ei, Ef)
}
return sum
}
fun getTotalLossDerivBivariateFunction(X: Double): BivariateFunction {
return BivariateFunction { Ei: Double, Ef: Double -> getTotalLossDeriv(X, Ei, Ef) }
}
/**
* Значение полной функции потерь с учетом всех неисчезающих порядков
*
* @param x
* @param Ei
* @param Ef
* @return
*/
fun getTotalLossValue(x: Double, Ei: Double, Ef: Double): Double {
if (x == 0.0) {
return 0.0
} else {
val probs = getLossProbabilities(x)
return (0 until probs.size).sumByDouble { i ->
probs[i] * getLossValue(i, Ei, Ef)
}
}
}
/**
* порог по вероятности, до которого вычисляются компоненты функции потерь
*/
private const val SCATTERING_PROBABILITY_THRESHOLD = 1e-3
private val integrator = GaussRuleIntegrator(100)
private val lossProbCache = Misc.getLRUCache<Double, List<Double>>(100)
private val A1 = 0.204
private val A2 = 0.0556
private val b = 14.0
private val pos1 = 12.6
private val pos2 = 14.3
private val w1 = 1.85
private val w2 = 12.5
val singleScatterFunction = UnivariateFunction { eps: Double ->
when {
eps <= 0 -> 0.0
eps <= b -> {
val z = eps - pos1
A1 * exp(-2.0 * z * z / w1 / w1)
}
else -> {
val z = 4.0 * (eps - pos2) * (eps - pos2)
A2 / (1 + z / w2 / w2)
}
}
}
/**
* A generic loss function for numass experiment in "Lobashev"
* parameterization
*
* @param exPos
* @param ionPos
* @param exW
* @param ionW
* @param exIonRatio
* @return
*/
fun getSingleScatterFunction(
exPos: Double,
ionPos: Double,
exW: Double,
ionW: Double,
exIonRatio: Double): UnivariateFunction {
val func = UnivariateFunction { eps: Double ->
if (eps <= 0) {
0.0
} else {
val z1 = eps - exPos
val ex = exIonRatio * exp(-2.0 * z1 * z1 / exW / exW)
val z = 4.0 * (eps - ionPos) * (eps - ionPos)
val ion = 1 / (1 + z / ionW / ionW)
if (eps < exPos) {
ex
} else {
Math.max(ex, ion)
}
}
}
val cutoff = 25.0
//caclulating lorentz integral analythically
val tailNorm = (Math.atan((ionPos - cutoff) * 2.0 / ionW) + 0.5 * Math.PI) * ionW / 2.0
val norm = integrator.integrate(0.0, cutoff, func)!! + tailNorm
return UnivariateFunction { e -> func.value(e) / norm }
}
fun getSingleScatterFunction(set: Values): UnivariateFunction {
val exPos = set.getDouble("exPos")
val ionPos = set.getDouble("ionPos")
val exW = set.getDouble("exW")
val ionW = set.getDouble("ionW")
val exIonRatio = set.getDouble("exIonRatio")
return getSingleScatterFunction(exPos, ionPos, exW, ionW, exIonRatio)
}
val trapFunction: BivariateFunction
get() = BivariateFunction { Ei: Double, Ef: Double ->
val eps = Ei - Ef
if (eps > 10) {
1.86e-04 * exp(-eps / 25.0) + 5.5e-05
} else {
0.0
}
}
fun plotScatter(frame: PlotFrame, set: Values) {
//"X", "shift", "exPos", "ionPos", "exW", "ionW", "exIonRatio"
// JFreeChartFrame frame = JFreeChartFrame.drawFrame("Differential scattering crosssection", null);
val X = set.getDouble("X")
val exPos = set.getDouble("exPos")
val ionPos = set.getDouble("ionPos")
val exW = set.getDouble("exW")
val ionW = set.getDouble("ionW")
val exIonRatio = set.getDouble("exIonRatio")
val scatterFunction = getSingleScatterFunction(exPos, ionPos, exW, ionW, exIonRatio)
if (set.names.contains("X")) {
val probs = LossCalculator.getGunLossProbabilities(set.getDouble("X"))
val single = { e: Double -> probs[1] * scatterFunction.value(e) }
frame.add(XYFunctionPlot.plot("Single scattering", 0.0, 100.0, 1000) { x: Double -> single(x) })
for (i in 2 until probs.size) {
val scatter = { e: Double -> probs[i] * LossCalculator.getLossValue(i, e, 0.0) }
frame.add(XYFunctionPlot.plot(i.toString() + " scattering", 0.0, 100.0, 1000) { x: Double -> scatter(x) })
}
val total = UnivariateFunction { eps ->
if (probs.size == 1) {
return@UnivariateFunction 0.0
}
var sum = probs[1] * scatterFunction.value(eps)
for (i in 2 until probs.size) {
sum += probs[i] * LossCalculator.getLossValue(i, eps, 0.0)
}
return@UnivariateFunction sum
}
frame.add(XYFunctionPlot.plot("Total loss", 0.0, 100.0, 1000) { x: Double -> total.value(x) })
} else {
frame.add(XYFunctionPlot.plot("Differential cross-section", 0.0, 100.0, 2000) { x: Double -> scatterFunction.value(x) })
}
}
}

View File

@ -66,7 +66,7 @@ fun main(args: Array<String>) {
for (hv in arrayOf(14000.0, 14500.0, 15000.0, 15500.0, 16050.0)) {
val frame = displayChart("integral[$hv]").apply {
this.plots.descriptor = Descriptors.forObject(DataPlot::class)
this.plots.descriptor = Descriptors.forType(DataPlot::class)
this.plots.configureValue("showLine", true)
}

View File

@ -47,7 +47,7 @@ fun main(args: Array<String>) {
val point = ProtoNumassPoint.readFile(Paths.get("D:\\Work\\Numass\\data\\2017_05_frames\\Fill_3_events\\set_33\\p36(30s)(HV1=17000).df"))
val frame = displayChart("integral").apply {
this.plots.descriptor = Descriptors.forObject(DataPlot::class)
this.plots.descriptor = Descriptors.forType(DataPlot::class)
this.plots.configureValue("showLine", true)
}

View File

@ -1,5 +1,5 @@
/*
* Copyright 2017 Alexander Nozik.
* Copyright 2018 Alexander Nozik.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -14,7 +14,7 @@
* limitations under the License.
*/
package inr.numass.scripts
package inr.numass.scripts.models
import hep.dataforge.description.Descriptors
import hep.dataforge.kodex.buildContext
@ -60,7 +60,7 @@ fun main(args: Array<String>) {
}
val frame = displayChart("differential").apply {
this.plots.descriptor = Descriptors.forObject(DataPlot::class)
this.plots.descriptor = Descriptors.forType(DataPlot::class)
this.plots.configureValue("showLine", true)
}

View File

@ -17,6 +17,7 @@
package inr.numass.scripts.models
import hep.dataforge.fx.output.FXOutputManager
import hep.dataforge.io.output.stream
import hep.dataforge.kodex.buildContext
import hep.dataforge.kodex.configure
import hep.dataforge.kodex.step
@ -25,10 +26,19 @@ import hep.dataforge.plots.Plot
import hep.dataforge.plots.data.DataPlot
import hep.dataforge.plots.jfreechart.JFreeChartPlugin
import hep.dataforge.plots.output.plot
import hep.dataforge.stat.fit.FitManager
import hep.dataforge.stat.fit.FitStage
import hep.dataforge.stat.fit.FitState
import hep.dataforge.stat.fit.ParamSet
import hep.dataforge.stat.models.XYModel
import hep.dataforge.tables.Adapters
import hep.dataforge.tables.ListTable
import inr.numass.NumassPlugin
import inr.numass.data.SpectrumAdapter
import inr.numass.models.NBkgSpectrum
import inr.numass.models.sterile.SterileNeutrinoSpectrum
import kotlinx.coroutines.experimental.launch
import java.io.PrintWriter
import kotlin.math.sqrt
@ -40,11 +50,9 @@ fun main(args: Array<String>) {
dataDir = "D:\\Work\\Numass\\data\\2018_04"
}
val sp = SterileNeutrinoSpectrum(context, Meta.empty())
//beta.setCaching(false);
val spectrum = NBkgSpectrum(SterileNeutrinoSpectrum(context, Meta.empty()))
val spectrum = NBkgSpectrum(sp)
//val model = XYModel(Meta.empty(), SpectrumAdapter(Meta.empty()), spectrum)
val t = 30 * 50 // time in seconds per point
val params = ParamSet().apply {
setPar("N", 8e5, 6.0, 0.0, Double.POSITIVE_INFINITY)
@ -57,9 +65,9 @@ fun main(args: Array<String>) {
setPar("trap", 1.0, 0.01)
}
fun ParamSet.update(vararg override: Pair<String, Double>): ParamSet = this.copy().apply {
fun ParamSet.update(vararg override: Pair<String, Double>): ParamSet = this.copy().also { set ->
override.forEach {
setParValue(it.first, it.second)
set.setParValue(it.first, it.second)
}
}
@ -70,19 +78,52 @@ fun main(args: Array<String>) {
}
fun plotResidual(name: String, vararg override: Pair<String, Double>): Plot {
val paramsMod = params.update(*override)
val x = (14000.0..18600.0).step(100.0).toList()
val y = x.map {
val base = spectrum.value(it, params)
val mod = spectrum.value(it, params.update(*override))
val err = sqrt(base/1e6)
return@map (mod - base) / err
val mod = spectrum.value(it, paramsMod)
val err = sqrt(base / t)
(mod - base) / err
}
return DataPlot.plot(name, x.toDoubleArray(), y.toDoubleArray())
}
val adapter = SpectrumAdapter(Meta.empty())
val fm = context.get<FitManager>()
fun plotFitResidual(name: String, vararg override: Pair<String, Double>): Plot {
val paramsMod = params.update(*override)
val x = (14000.0..18400.0).step(100.0).toList()
val table = ListTable.Builder(Adapters.getFormat(adapter)).apply {
x.forEach { u ->
row(adapter.buildSpectrumDataPoint(u, t * spectrum.value(u, params).toLong(), t.toDouble()))
}
}.build()
val model = XYModel(Meta.empty(), adapter, spectrum)
val state = FitState(table, model, params)
val res = fm.runStage(state, "QOW", FitStage.TASK_RUN, "N", "E0","bkg")
res.printState(PrintWriter(System.out))
res.printState(PrintWriter(context.output["fitResult", name].stream))
//context.output["fitResult",name].stream
val y = x.map { u ->
val base = spectrum.value(u, params)
val mod = spectrum.value(u, res.parameters)
val err = sqrt(base / t)
(mod - base) / err
}
return DataPlot.plot(name, x.toDoubleArray(), y.toDoubleArray())
}
context.plot("default") {
/*
context.plot("trap") {
plots.configure {
"showLine" to true
"showSymbol" to false
@ -100,10 +141,37 @@ fun main(args: Array<String>) {
"showErrors" to false
}
plots.setType<DataPlot>()
+plotResidual("sterile_1","U2" to 1e-3)
+plotResidual("sterile_3","msterile2" to (3000*3000).toDouble(),"U2" to 1e-3)
+plotResidual("X","X" to 0.11)
+plotResidual("sterile_1", "U2" to 1e-3)
+plotResidual("sterile_3", "msterile2" to (3000 * 3000).toDouble(), "U2" to 1e-3)
+plotResidual("X", "X" to 0.11)
+plotResidual("trap", "trap" to 0.99)
}*/
context.plot("fit", stage = "plots") {
plots.configure {
"showLine" to true
"showSymbol" to false
"showErrors" to false
"thickness" to 4.0
}
plots.setType<DataPlot>()
launch {
+plotResidual("trap", "trap" to 0.99)
+plotFitResidual("trap_fit", "trap" to 0.99)
}
launch {
+plotResidual("X", "X" to 0.11)
+plotFitResidual("X_fit", "X" to 0.11)
}
launch {
+plotResidual("sterile_1", "U2" to 1e-3)
+plotFitResidual("sterile_1_fit", "U2" to 1e-3)
}
launch {
+plotResidual("sterile_3", "msterile2" to (3000 * 3000).toDouble(), "U2" to 1e-3)
+plotFitResidual("sterile_3_fit", "msterile2" to (3000 * 3000).toDouble(), "U2" to 1e-3)
}
}
}