AutomaticDifferentiation/AutomaticDifferentiationVector.hpp

#ifndef DEF_AUTOMATIC_DIFFERENTIATION
#define DEF_AUTOMATIC_DIFFERENTIATION

#include <assert.h>
#include <cmath>
#include <ostream>

#include <valarray>

template<typename T>
std::ostream & operator<<(std::ostream & out, std::valarray<T> const& v)
{
	for(size_t i = 0 ; i < v.size() ; i++)
		out << v[i] << " ";
	return out;
}

#define MINAB(a, b) (((a) < (b)) ? (a) : (b))

/// Implementation of dual numbers for automatic differentiation.
/// This implementation uses vectors for b so that function gradients can be computed in one function call.
/// Set the index of every variable with the ::d(int i) function and call the function to be computed : f(x+DualVector::d(0), y+DualVector::d(1), z+DualVector::d(2), ...)
/// reference : http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.89.7749&rep=rep1&type=pdf
template<typename Scalar>
class DualVector
{
	public:
		using VectorT = std::valarray<Scalar>;
		
		static VectorT __create_VectorT_zeros(int N = 1)
		{
			assert(N >= 0);
			VectorT res(Scalar(0.), N);
			return res;
		}

		DualVector(const Scalar & _a = Scalar(0.), const VectorT & _b = DualVector::__create_VectorT_zeros())
			: a(_a),
			  b(_b)
			  {}
		
		DualVector const& operator=(Scalar const& _a)
		{
			*this = DualVector(_a);
		}

		/// Use this function to set what variable is to be derived : x + DualVector::d(i)
		static DualVector D(int i = 0, int N = 1)
		{
			assert(i >= 0);
			assert(i < N);
			VectorT _d = DualVector::__create_VectorT_zeros(N);
			_d[i] = Scalar(1.);
			return DualVector(Scalar(0.), _d);
		}

		/// Use this function to set what variable is to be derived.
		DualVector const& diff(int i = 0, int N = 1)
		{
			assert(i >= 0);
			assert(i < N);
			if(N != b.size())
			{
				// copy old data into new b vector
				VectorT b_old = b;
				b.resize(N);
				for(size_t j = 0 ; j < MINAB(b.size(), b_old.size()) ; j++)
					b[j] = b_old[j];
			}
			b[i] = Scalar(1.);
			return *this;
		}
		
		/// Returns the value
		Scalar const& x() const
		{
			return a;
		}

		/// Returns the value
		Scalar & x()
		{
			return a;
		}

		/// Returns the derivative value at index i
		Scalar const& d(int i) const
		{
			assert(i >= 0);
			assert(i < b.size());
			return b[i];
		}

		/// Returns the derivative value at index i
		Scalar & d(int i)
		{
			assert(i >= 0);
			assert(i < b.size());
			return b[i];
		}

		DualVector & operator+=(const DualVector & x)
		{
			a += x.a;
			b += x.b;
			return *this;
		}
		
		DualVector & operator-=(const DualVector & x)
		{
			a -= x.a;
			b -= x.b;
			return *this;
		}
		
		DualVector & operator*=(const DualVector & x)
		{
			b = a*x.b + b*x.a;
			a *= x.a;
			return *this;
		}
		
		DualVector & operator/=(const DualVector & x)
		{
			b = (x.a*b - a*x.b)/(x.a*x.a);
			a /= x.a;
			return *this;
		}
		
		DualVector & operator++() { // ++x
			return ((*this) += Scalar(1.));
		}
		
		DualVector & operator--() { // --x
			return ((*this) -= Scalar(1.));
		}
		
		DualVector operator++(int) { // x++
			DualVector copy = *this;
			(*this) += Scalar(1.);
			return copy;
		}
		
		DualVector operator--(int) { // x--
			DualVector copy = *this;
			(*this) -= Scalar(1.);
			return copy;
		}
		
		DualVector operator+(const DualVector & x) const {
			DualVector res(*this);
			return (res += x);
		}
		
		DualVector operator+(void) const // +x
		{
			return (*this);
		}
		
		DualVector operator-(const DualVector & x) const {
			DualVector res(*this);
			return (res -= x);
		}
		
		DualVector operator-(void) const // -x
		{
			return DualVector(-a, -b);
		}
		
		DualVector operator*(const DualVector & x) const
		{
			DualVector res(*this);
			return (res *= x);
		}
		
		DualVector operator/(const DualVector & x) const
		{
			DualVector res(*this);
			return (res /= x);
		}
		
		bool operator==(const DualVector & x) const {
			return (a == x.a);
		}
		
		bool operator!=(const DualVector & x) const {
			return (a != x.a);
		}
		
		bool operator<(const DualVector & x) const {
			return (a < x.a);
		}
		
		bool operator<=(const DualVector & x) const {
			return (a <= x.a);
		}
		
		bool operator>(const DualVector & x) const {
			return (a > x.a);
		}
		
		bool operator>=(const DualVector & x) const {
			return (a >= x.a);
		}
		
		Scalar a;	/// Real part
		VectorT b;	/// Infinitesimal parts
};

template<typename A, typename B>
DualVector<B> operator+(A const& v, DualVector<B> const& x) {
	return (DualVector<B>(v) + x);
}
template<typename A, typename B>
DualVector<B> operator-(A const& v, DualVector<B> const& x) {
	return (DualVector<B>(v) - x);
}
template<typename A, typename B>
DualVector<B> operator*(A const& v, DualVector<B> const& x) {
	return (DualVector<B>(v) * x);
}
template<typename A, typename B>
DualVector<B> operator/(A const& v, DualVector<B> const& x) {
	return (DualVector<B>(v) / x);
}

// Basic mathematical functions for Scalar numbers

// Trigonometric functions
template<typename Scalar> Scalar sec(const Scalar & x) {
	return Scalar(1.)/cos(x);
}

template<typename Scalar> Scalar cot(const Scalar & x) {
	return cos(x)/sin(x);
}

template<typename Scalar> Scalar csc(const Scalar & x) {
	return Scalar(1.)/sin(x);
}

// Inverse trigonometric functions
template<typename Scalar> Scalar asec(const Scalar & x) {
	return acos(Scalar(1.)/x);
}

template<typename Scalar> Scalar acot(const Scalar & x) {
	return atan(Scalar(1.)/x);
}

template<typename Scalar> Scalar acsc(const Scalar & x) {
	return asin(Scalar(1.)/x);
}

// Hyperbolic trigonometric functions
template<typename Scalar> Scalar sech(const Scalar & x) {
	return Scalar(1.)/cosh(x);
}

template<typename Scalar> Scalar coth(const Scalar & x) {
	return cosh(x)/sinh(x);
}

template<typename Scalar> Scalar csch(const Scalar & x) {
	return Scalar(1.)/sinh(x);
}

// Inverse hyperbolic trigonometric functions
template<typename Scalar> Scalar asech(const Scalar & x) {
	return log((Scalar(1.) + sqrt(Scalar(1.) - x*x))/x);
}

template<typename Scalar> Scalar acoth(const Scalar & x) {
	return Scalar(0.5)*log((x + Scalar(1.))/(x - Scalar(1.)));
}

template<typename Scalar> Scalar acsch(const Scalar & x) {
	return (x >= Scalar(0.)) ? log((Scalar(1.) + sqrt(Scalar(1.) + x*x))/x) : log((Scalar(1.) - sqrt(Scalar(1.) + x*x))/x);
}

// Other functions
template<typename Scalar> Scalar exp10(const Scalar & x) {
	return exp(x*log(Scalar(10.)));
}

template<typename Scalar> Scalar sign(const Scalar & x) {
	return (x >= Scalar(0.)) ? ((x > Scalar(0.)) ? Scalar(1.) : Scalar(0.)) : Scalar(-1.);
}

template<typename Scalar> Scalar heaviside(const Scalar & x) {
	return Scalar(x >= Scalar(0.));
}

template<typename Scalar> Scalar abs(const Scalar & x) {
	return (x >= Scalar(0.)) ? x : -x;
}

// Basic mathematical functions for DualVector numbers
// f(a + b*d) = f(a) + b*f'(a)*d

// Trigonometric functions
template<typename Scalar> DualVector<Scalar> cos(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(cos(x.a), -x.b*sin(x.a));
}

template<typename Scalar> DualVector<Scalar> sin(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(sin(x.a), x.b*cos(x.a));
}

template<typename Scalar> DualVector<Scalar> tan(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(tan(x.a), x.b*sec(x.a)*sec(x.a));
}

template<typename Scalar> DualVector<Scalar> sec(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(sec(x.a), x.b*sec(x.a)*tan(x.a));
}

template<typename Scalar> DualVector<Scalar> cot(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(cot(x.a), x.b*(-csc(x.a)*csc(x.a)));
}

template<typename Scalar> DualVector<Scalar> csc(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(csc(x.a), x.b*(-cot(x.a)*csc(x.a)));
}

// Inverse trigonometric functions
template<typename Scalar> DualVector<Scalar> acos(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(acos(x.a), x.b*(-Scalar(1.)/sqrt(Scalar(1.)-x.a*x.a)));
}

template<typename Scalar> DualVector<Scalar> asin(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(asin(x.a), x.b*(Scalar(1.)/sqrt(Scalar(1.)-x.a*x.a)));
}

template<typename Scalar> DualVector<Scalar> atan(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(atan(x.a), x.b*(Scalar(1.)/(x.a*x.a + Scalar(1.))));
}

template<typename Scalar> DualVector<Scalar> asec(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(asec(x.a), x.b*(Scalar(1.)/(sqrt(Scalar(1.)-Scalar(1.)/(x.a*x.a))*(x.a*x.a))));
}

template<typename Scalar> DualVector<Scalar> acot(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(acot(x.a), x.b*(-Scalar(1.)/((x.a*x.a)+Scalar(1.))));
}

template<typename Scalar> DualVector<Scalar> acsc(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(acsc(x.a), x.b*(-Scalar(1.)/(sqrt(Scalar(1.)-Scalar(1.)/(x.a*x.a))*(x.a*x.a))));
}

// Hyperbolic trigonometric functions
template<typename Scalar> DualVector<Scalar> cosh(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(cosh(x.a), x.b*sinh(x.a));
}

template<typename Scalar> DualVector<Scalar> sinh(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(sinh(x.a), x.b*cosh(x.a));
}

template<typename Scalar> DualVector<Scalar> tanh(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(tanh(x.a), x.b*sech(x.a)*sech(x.a));
}

template<typename Scalar> DualVector<Scalar> sech(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(sech(x.a), x.b*(-sech(x.a)*tanh(x.a)));
}

template<typename Scalar> DualVector<Scalar> coth(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(coth(x.a), x.b*(-csch(x.a)*csch(x.a)));
}

template<typename Scalar> DualVector<Scalar> csch(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(csch(x.a), x.b*(-coth(x.a)*csch(x.a)));
}

// Inverse hyperbolic trigonometric functions
template<typename Scalar> DualVector<Scalar> acosh(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(acosh(x.a), x.b*(Scalar(1.)/sqrt((x.a*x.a)-Scalar(1.))));
}

template<typename Scalar> DualVector<Scalar> asinh(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(asinh(x.a), x.b*(Scalar(1.)/sqrt((x.a*x.a)+Scalar(1.))));
}

template<typename Scalar> DualVector<Scalar> atanh(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(atanh(x.a), x.b*(Scalar(1.)/(Scalar(1.)-(x.a*x.a))));
}

template<typename Scalar> DualVector<Scalar> asech(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(asech(x.a), x.b*(Scalar(-1.)/(sqrt(Scalar(1.)/(x.a*x.a)-Scalar(1.))*(x.a*x.a))));
}

template<typename Scalar> DualVector<Scalar> acoth(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(acoth(x.a), x.b*(-Scalar(1.)/((x.a*x.a)-Scalar(1.))));
}

template<typename Scalar> DualVector<Scalar> acsch(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(acsch(x.a), x.b*(-Scalar(1.)/(sqrt(Scalar(1.)/(x.a*x.a)+Scalar(1.))*(x.a*x.a))));
}

// Exponential functions
template<typename Scalar> DualVector<Scalar> exp(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(exp(x.a), x.b*exp(x.a));
}

template<typename Scalar> DualVector<Scalar> log(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(log(x.a), x.b/x.a);
}

template<typename Scalar> DualVector<Scalar> exp10(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(exp10(x.a), x.b*(log(Scalar(10.))*exp10(x.a)));
}

template<typename Scalar> DualVector<Scalar> log10(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(log10(x.a), x.b/(log(Scalar(10.))*x.a));
}

template<typename Scalar> DualVector<Scalar> exp2(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(exp2(x.a), x.b*(log(Scalar(2.))*exp2(x.a)));
}

template<typename Scalar> DualVector<Scalar> log2(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(log2(x.a), x.b/(log(Scalar(2.))*x.a));
}

template<typename Scalar> DualVector<Scalar> pow(const DualVector<Scalar> & x, const DualVector<Scalar> & n) {
	return exp(n*log(x));
}

// Other functions
template<typename Scalar> DualVector<Scalar> sqrt(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(sqrt(x.a), x.b/(Scalar(2.)*sqrt(x.a)));
}

template<typename Scalar> DualVector<Scalar> sign(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(sign(x.a), DualVector<Scalar>::DualVector::__create_VectorT_zeros());
}

template<typename Scalar> DualVector<Scalar> abs(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(abs(x.a), x.b*sign(x.a));
}
template<typename Scalar> DualVector<Scalar> fabs(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(fabs(x.a), x.b*sign(x.a));
}

template<typename Scalar> DualVector<Scalar> heaviside(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(heaviside(x.a), DualVector<Scalar>::DualVector::__create_VectorT_zeros());
}

template<typename Scalar> DualVector<Scalar> floor(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(floor(x.a), DualVector<Scalar>::DualVector::__create_VectorT_zeros());
}

template<typename Scalar> DualVector<Scalar> ceil(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(ceil(x.a), DualVector<Scalar>::DualVector::__create_VectorT_zeros());
}

template<typename Scalar> DualVector<Scalar> round(const DualVector<Scalar> & x) {
	return DualVector<Scalar>(round(x.a), DualVector<Scalar>::DualVector::__create_VectorT_zeros());
}

template<typename Scalar> std::ostream & operator<<(std::ostream & s, const DualVector<Scalar> & x)
{
	return (s << x.a);
}

#endif