-
-
Notifications
You must be signed in to change notification settings - Fork 7.3k
/
Copy pathfast_fourier_transform.cpp
167 lines (140 loc) · 5.24 KB
/
fast_fourier_transform.cpp
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
/**
* @file
* @brief [A fast Fourier transform
* (FFT)](https://medium.com/@aiswaryamathur/understanding-fast-fouriertransform-from-scratch-to-solve-polynomial-multiplication-8018d511162f)
* is an algorithm that computes the
* discrete Fourier transform (DFT) of a sequence, or its inverse (IDFT).
* @details
* This
* algorithm has application in use case scenario where a user wants to find
points of a
* function
* in a short time by just using the coefficients of the polynomial
* function.
* It can be also used to find inverse fourier transform by just switching the
value of omega.
* Time complexity
* this algorithm computes the DFT in O(nlogn) time in comparison to traditional
O(n^2).
* @author [Ameya Chawla](https://github.com/ameyachawlaggsipu)
*/
#include <cassert> /// for assert
#include <cmath> /// for mathematical-related functions
#include <complex> /// for storing points and coefficents
#include <cstdint>
#include <iostream> /// for IO operations
#include <vector> /// for std::vector
/**
* @namespace numerical_methods
* @brief Numerical algorithms/methods
*/
namespace numerical_methods {
/**
* @brief FastFourierTransform 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<double> *FastFourierTransform(std::complex<double> *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<double> om = std::complex<double>(
cos(2 * pi / n), sin(2 * pi / n)); /// Calculating value of omega
auto *pe = new std::complex<double>[n / 2]; /// Coefficients of even power
auto *po = new std::complex<double>[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<double> *ye =
FastFourierTransform(pe, n / 2); /// Recursive Call
std::complex<double> *yo =
FastFourierTransform(po, n / 2); /// Recursive Call
auto *y = new std::complex<double>[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<double>[2]; /// Test case 1
auto *t2 = new std::complex<double>[4]; /// Test case 2
t1[0] = {1, 0};
t1[1] = {2, 0};
t2[0] = {1, 0};
t2[1] = {2, 0};
t2[2] = {3, 0};
t2[3] = {4, 0};
uint8_t n1 = 2;
uint8_t n2 = 4;
std::vector<std::complex<double>> r1 = {
{3, 0}, {-1, 0}}; /// True Answer for test case 1
std::vector<std::complex<double>> r2 = {
{10, 0}, {-2, -2}, {-2, 0}, {-2, 2}}; /// True Answer for test case 2
std::complex<double> *o1 = numerical_methods::FastFourierTransform(t1, n1);
std::complex<double> *t3 =
o1; /// Temporary variable used to delete memory location of o1
std::complex<double> *o2 = numerical_methods::FastFourierTransform(t2, n2);
std::complex<double> *t4 =
o2; /// Temporary variable used to delete memory location of o2
for (uint8_t i = 0; i < n1; i++) {
assert((r1[i].real() - o1->real() < 0.000000000001) &&
(r1[i].imag() - o1->imag() <
0.000000000001)); /// Comparing for both real and imaginary
/// values for test case 1
o1++;
}
for (uint8_t i = 0; i < n2; i++) {
assert((r2[i].real() - o2->real() < 0.000000000001) &&
(r2[i].imag() - o2->imag() <
0.000000000001)); /// Comparing for both real and imaginary
/// values for test case 2
o2++;
}
delete[] t1;
delete[] t2;
delete[] t3;
delete[] t4;
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;
}