/** * @file * @brief [An inverse fast Fourier transform * (IFFT)](https://www.geeksforgeeks.org/python-inverse-fast-fourier-transformation/) * is an algorithm that computes the inverse fourier transform. * @details * This algorithm has an application in use case scenario where a user wants * find coefficients of a function in a short time by just using points * generated by DFT. Time complexity this algorithm computes the IDFT in * O(nlogn) time in comparison to traditional O(n^2). * @author [Ameya Chawla](https://github.com/ameyachawlaggsipu) */ #include /// for assert #include /// for mathematical-related functions #include /// for storing points and coefficents #include /// for IO operations #include /// for std::vector /** * @namespace numerical_methods * @brief Numerical algorithms/methods */ namespace numerical_methods { /** * @brief InverseFastFourierTransform is a recursive function which returns list * of complex numbers * @param p List of Coefficents in form of complex numbers * @param n Count of elements in list p * @returns p if n==1 * @returns y if n!=1 */ std::complex *InverseFastFourierTransform(std::complex *p, uint8_t n) { if (n == 1) { return p; /// Base Case To return } double pi = 2 * asin(1.0); /// Declaring value of pi std::complex om = std::complex( cos(2 * pi / n), sin(2 * pi / n)); /// Calculating value of omega om.real(om.real() / n); /// One change in comparison with DFT om.imag(om.imag() / n); /// One change in comparison with DFT auto *pe = new std::complex[n / 2]; /// Coefficients of even power auto *po = new std::complex[n / 2]; /// Coefficients of odd power int k1 = 0, k2 = 0; for (int j = 0; j < n; j++) { if (j % 2 == 0) { pe[k1++] = p[j]; /// Assigning values of even Coefficients } else { po[k2++] = p[j]; /// Assigning value of odd Coefficients } } std::complex *ye = InverseFastFourierTransform(pe, n / 2); /// Recursive Call std::complex *yo = InverseFastFourierTransform(po, n / 2); /// Recursive Call auto *y = new std::complex[n]; /// Final value representation list k1 = 0, k2 = 0; for (int i = 0; i < n / 2; i++) { y[i] = ye[k1] + pow(om, i) * yo[k2]; /// Updating the first n/2 elements y[i + n / 2] = ye[k1] - pow(om, i) * yo[k2]; /// Updating the last n/2 elements k1++; k2++; } if (n != 2) { delete[] pe; delete[] po; } delete[] ye; /// Deleting dynamic array ye delete[] yo; /// Deleting dynamic array yo return y; } } // namespace numerical_methods /** * @brief Self-test implementations * @details * Declaring two test cases and checking for the error * in predicted and true value is less than 0.000000000001. * @returns void */ static void test() { /* descriptions of the following test */ auto *t1 = new std::complex[2]; /// Test case 1 auto *t2 = new std::complex[4]; /// Test case 2 t1[0] = {3, 0}; t1[1] = {-1, 0}; t2[0] = {10, 0}; t2[1] = {-2, -2}; t2[2] = {-2, 0}; t2[3] = {-2, 2}; uint8_t n1 = 2; uint8_t n2 = 4; std::vector> r1 = { {1, 0}, {2, 0}}; /// True Answer for test case 1 std::vector> r2 = { {1, 0}, {2, 0}, {3, 0}, {4, 0}}; /// True Answer for test case 2 std::complex *o1 = numerical_methods::InverseFastFourierTransform(t1, n1); std::complex *o2 = numerical_methods::InverseFastFourierTransform(t2, n2); for (uint8_t i = 0; i < n1; i++) { assert((r1[i].real() - o1[i].real() < 0.000000000001) && (r1[i].imag() - o1[i].imag() < 0.000000000001)); /// Comparing for both real and imaginary /// values for test case 1 } for (uint8_t i = 0; i < n2; i++) { assert((r2[i].real() - o2[i].real() < 0.000000000001) && (r2[i].imag() - o2[i].imag() < 0.000000000001)); /// Comparing for both real and imaginary /// values for test case 2 } delete[] t1; delete[] t2; delete[] o1; delete[] o2; std::cout << "All tests have successfully passed!\n"; } /** * @brief Main function * @param argc commandline argument count (ignored) * @param argv commandline array of arguments (ignored) * calls automated test function to test the working of fast fourier transform. * @returns 0 on exit */ int main(int argc, char const *argv[]) { test(); // run self-test implementations // with 2 defined test cases return 0; }