Upsilon/apps/probability/law/exponential_law.cpp

#include "exponential_law.h"
#include <assert.h>
#include <math.h>
#include <float.h>
#include <ion.h>

namespace Probability {

ExponentialLaw::ExponentialLaw() :
  OneParameterLaw(1.0f)
{
}

const char * ExponentialLaw::title() {
  return "Loi exponentielle";
}

Law::Type ExponentialLaw::type() const {
  return Type::Exponential;
}

bool ExponentialLaw::isContinuous() const {
  return true;
}

const char * ExponentialLaw::parameterNameAtIndex(int index) {
  assert(index == 0);
  constexpr static char name[] = {Ion::Charset::SmallLambda, 0};
  return name;
}

const char * ExponentialLaw::parameterDefinitionAtIndex(int index) {
  assert(index == 0);
  constexpr static char def[] = {Ion::Charset::SmallLambda, ' ', ':', ' ', 'p', 'a', 'r', 'a', 'm', 'e', 't', 'r', 'e', 0};
  return def;
}

float ExponentialLaw::xMin() {
  float max = xMax();
  return - k_displayLeftMarginRatio * max;
}

float ExponentialLaw::xMax() {
  assert(m_parameter1 != 0.0f);
  float result = 5.0f/m_parameter1;
  if (result <= 0.0f) {
    result = 1.0f;
  }
  return result*(1.0f+ k_displayRightMarginRatio);
}

float ExponentialLaw::yMin() {
  return -k_displayBottomMarginRatio*yMax();
}

float ExponentialLaw::yMax() {
  float result = m_parameter1;
  if (result <= 0.0f || isnan(result)) {
    result = 1.0f;
  }
  if (result <= 0.0f) {
    result = 1.0f;
  }
  return result*(1.0f+ k_displayTopMarginRatio);
}

float ExponentialLaw::evaluateAtAbscissa(float x) const {
  if (x < 0.0f) {
    return NAN;
  }
  return m_parameter1*expf(-m_parameter1*x);
}

bool ExponentialLaw::authorizedValueAtIndex(float x, int index) const {
  if (x <= 0.0f) {
    return false;
  }
  return true;
}

float ExponentialLaw::cumulativeDistributiveFunctionAtAbscissa(float x) const {
  return 1.0f - expf(-m_parameter1*x);
}

float ExponentialLaw::cumulativeDistributiveInverseForProbability(float * probability) {
  if (*probability >= 1.0f) {
    return INFINITY;
  }
  if (*probability <= 0.0f) {
    return 0.0f;
  }
  return -logf(1.0f - *probability)/m_parameter1;
}

}