Table of Contents
Matrix normal
algorithm
Example
Output
Matrix trace
in conclusion
Home Backend Development C++ C++ program to find normal vectors and traces

C++ program to find normal vectors and traces

Sep 09, 2023 pm 04:41 PM
c program trace Find normal vector

C++ program to find normal vectors and traces

Two-dimensional arrays or matrices are very useful in several applications. A matrix has rows and columns and stores numbers in it. In C, we can also use multidimensional arrays to define two-dimensional matrices. In this article, we will see how to calculate the norm and trace of a given matrix using C.

The normal is the square root of the sum of all elements in the matrix. The trace is the sum of the elements present in the main diagonal. Let's look at the algorithm and C code representation.

Matrix normal

$\begin{bmatrix} 5 & ​​1& 8\line break 4 & 3& 9\line break 2&7&3\ \end{bmatrix},$

Sum of all elements: (5 + 1 + 8 + 4 + 3 + 9 + 2 + 7 + 3) = 42
Normal: (Square root of the sum of all elements) = √42 = 6.48
Copy after login

In the above example, we took a 3 x 3 matrix, here we get the sum of all elements and then take the square root of it. Let us see the algorithm for better understanding.

algorithm

  • Read matrix M as input
  • Consider M with n rows and n columns
  • Sum: = 0
  • For i from 1 to n, execute
    • For j from 1 to n, perform the following operations
      • sum := sum M[ i ][ j ]
    • End loop
  • End loop
  • res := sum of square roots
  • Return results

Example

#include <iostream>
#include <cmath>
#define N 5
using namespace std;
float solve( int M[ N ][ N ] ){
   int sum = 0;
   for ( int i = 0; i < N; i++ ) {
      for ( int j = 0; j < N; j++ ) {
         sum = sum + M[ i ][ j ];
      }
   }
   return sqrt( sum );
}
int main(){
   int mat1[ N ][ N ] = {
      {5, 8, 74, 21, 69},
      {48, 2, 98, 6, 63},
      {85, 12, 10, 6, 9},
      {6, 12, 18, 32, 5},
      {8, 45, 74, 69, 1},
   };
   cout << "Normal of the first matrix is: " << solve( mat1 ) << endl;
   int mat2[ N ][ N ] = {
      {6, 8, 35, 21, 87},
      {99, 2, 36, 326, 25},
      {15, 215, 3, 157, 8},
      {96, 115, 17, 5, 3},
      {56, 4, 78, 5, 10},
   };
   cout << "Normal of the second matrix is: " << solve( mat2 ) << endl;
}
Copy after login

Output

Normal of the first matrix is: 28.0357
Normal of the second matrix is: 37.8418
Copy after login

Matrix trace

$\begin{bmatrix} 5 & ​​1& 8\line break 4 & 3& 9\line break 2&7&3\ \end{bmatrix},$

Sum of all elements in main diagonal: (5 + 3 + 3) = 11 which is
the trace of given matrix
Copy after login

In the above example, we took a 3 x 3 matrix, where we got the sum of all the elements on the main diagonal. This sum is the trace of the matrix. Let’s take a look at the algorithm for better understanding.

algorithm

  • Read matrix M as input
  • Consider M with n rows and n columns
  • Sum: = 0
  • For i from 1 to n, execute
    • sum := sum M[ i ][ i ]
  • End loop
  • Return the sum

Example

#include <iostream>
#include <cmath>
#define N 5
using namespace std;
float solve( int M[ N ][ N ] ){
   int sum = 0;
   for ( int i = 0; i < N; i++ ) {
      sum = sum + M[ i ][ i ];
   }
   return sum;
}
int main(){
   int mat1[ N ][ N ] = {
      {5, 8, 74, 21, 69},
      {48, 2, 98, 6, 63},
      {85, 12, 10, 6, 9},
      {6, 12, 18, 32, 5},
      {8, 45, 74, 69, 1},
   };
   cout << "Trace of the first matrix is: " << solve( mat1 ) << endl;
   int mat2[ N ][ N ] = {
      {6, 8, 35, 21, 87},
      {99, 2, 36, 326, 25},
      {15, 215, 3, 157, 8},
      {96, 115, 17, 5, 3},
      {56, 4, 78, 5, 10},
   };
   cout << "Trace of the second matrix is: " << solve( mat2 ) << endl;
}
Copy after login

Output

Trace of the first matrix is: 50
Trace of the second matrix is: 26
Copy after login

in conclusion

Normals and traces are both matrix operations. In order to perform these two operations, we need a square matrix (because a square trace matrix is ​​required). The normal is simply the square root of the sum of all elements present in the matrix, and the trace is the sum of the elements present on the main diagonal of the matrix. This matrix can be represented using a two-dimensional array in C. Here we give two examples of matrices with 5 rows and 5 columns (25 elements in total). Accessing the matrix requires a loop statement with indexing operations. For normal calculations we need to iterate through each element, so two nested loops are required. The complexity of this program is O(n2). For tracing, since we only need to look at the main diagonal, the row index and column index will be the same. So just one for loop is enough. Can be calculated in O(n) time.

The above is the detailed content of C++ program to find normal vectors and traces. For more information, please follow other related articles on the PHP Chinese website!

Statement of this Website
The content of this article is voluntarily contributed by netizens, and the copyright belongs to the original author. This site does not assume corresponding legal responsibility. If you find any content suspected of plagiarism or infringement, please contact admin@php.cn

Hot AI Tools

Undresser.AI Undress

Undresser.AI Undress

AI-powered app for creating realistic nude photos

AI Clothes Remover

AI Clothes Remover

Online AI tool for removing clothes from photos.

Undress AI Tool

Undress AI Tool

Undress images for free

Clothoff.io

Clothoff.io

AI clothes remover

Video Face Swap

Video Face Swap

Swap faces in any video effortlessly with our completely free AI face swap tool!

Hot Tools

Notepad++7.3.1

Notepad++7.3.1

Easy-to-use and free code editor

SublimeText3 Chinese version

SublimeText3 Chinese version

Chinese version, very easy to use

Zend Studio 13.0.1

Zend Studio 13.0.1

Powerful PHP integrated development environment

Dreamweaver CS6

Dreamweaver CS6

Visual web development tools

SublimeText3 Mac version

SublimeText3 Mac version

God-level code editing software (SublimeText3)

Translate the following into Chinese: C program to convert Roman numerals to decimal numbers Translate the following into Chinese: C program to convert Roman numerals to decimal numbers Sep 05, 2023 pm 09:53 PM

Given below is a C language algorithm to convert Roman numerals to decimal numbers: Algorithm Step 1 - Start Step 2 - Read Roman numerals at runtime Step 3 - Length: = strlen(roman) Step 4 - For i=0 to Length-1 Step 4.1-switch(roman[i]) Step 4.1.1-case'm': &nbs

C++ program to compare the lexicographic order of two strings C++ program to compare the lexicographic order of two strings Sep 04, 2023 pm 05:13 PM

Lexicographic string comparison means that strings are compared in dictionary order. For example, if there are two strings 'apple' and 'appeal', the first string will come last because the first three characters of 'app' are the same. Then for the first string the character is 'l' and in the second string the fourth character is 'e'. Since 'e' is shorter than 'l', it will come first if we sort lexicographically. Strings are compared lexicographically before being arranged. In this article, we will see different techniques for lexicographically comparing two strings using C++. Using the compare() function in C++ strings The C++string object has a compare()

C program to find length of linked list C program to find length of linked list Sep 07, 2023 pm 07:33 PM

Linked lists use dynamic memory allocation, i.e. they grow and shrink accordingly. They are defined as collections of nodes. Here, a node has two parts, data and links. The representation of data, links and linked lists is as follows - Types of linked lists There are four types of linked lists, as follows: - Single linked list/Singly linked list Double/Doubly linked list Circular single linked list Circular double linked list We use the recursive method to find the length of the linked list The logic is -intlength(node *temp){ if(temp==NULL) returnl; else{&n

C program uses rename() function to change file name C program uses rename() function to change file name Sep 21, 2023 pm 10:01 PM

The rename function changes a file or directory from its old name to its new name. This operation is similar to the move operation. So we can also use this rename function to move files. This function exists in the stdio.h library header file. The syntax of the rename function is as follows: intrename(constchar*oldname,constchar*newname); The function of the rename() function accepts two parameters. One is oldname and the other is newname. Both parameters are pointers to constant characters that define the old and new names of the file. Returns zero if the file was renamed successfully; otherwise, returns a nonzero integer. During a rename operation

C++ program to find the value of the inverse hyperbolic sine function taking a given value as argument C++ program to find the value of the inverse hyperbolic sine function taking a given value as argument Sep 17, 2023 am 10:49 AM

Hyperbolic functions are defined using hyperbolas instead of circles and are equivalent to ordinary trigonometric functions. It returns the ratio parameter in the hyperbolic sine function from the supplied angle in radians. But do the opposite, or in other words. If we want to calculate an angle from a hyperbolic sine, we need an inverse hyperbolic trigonometric operation like the hyperbolic inverse sine operation. This course will demonstrate how to use the hyperbolic inverse sine (asinh) function in C++ to calculate angles using the hyperbolic sine value in radians. The hyperbolic arcsine operation follows the following formula -$$\mathrm{sinh^{-1}x\:=\:In(x\:+\:\sqrt{x^2\:+\:1})}, Where\:In\:is\:natural logarithm\:(log_e\:k)

C++ program to print dictionary C++ program to print dictionary Sep 11, 2023 am 10:33 AM

A map is a special type of container in C++ in which each element is a pair of two values, namely a key value and a map value. The key value is used to index each item, and the mapped value is the value associated with the key. Regardless of whether the mapped value is unique, the key is always unique. To print map elements in C++ we have to use iterator. An element in a set of items is indicated by an iterator object. Iterators are primarily used with arrays and other types of containers (such as vectors), and they have a specific set of operations that can be used to identify specific elements within a specific range. Iterators can be incremented or decremented to reference different elements present in a range or container. The iterator points to the memory location of a specific element in the range. Printing a map in C++ using iterators First, let's look at how to define

C++ program to check if a character is alphabetic or non-alphabetic C++ program to check if a character is alphabetic or non-alphabetic Sep 14, 2023 pm 03:37 PM

Using strings or characters is sometimes very useful when solving some logic programming problems. A string is a collection of characters, which is a 1-byte data type used to hold symbols in ASCII values. Symbols can be English letters, numbers, or special characters. In this article, we will learn how to check if a character is an English letter or a letter of the alphabet using C++. Checking the isalpha() function To check if a number is a letter, we can use the isalpha() function in the ctype.h header file. This takes a character as input and returns true if it is an alphabet, false otherwise. Let us look at the following C++ implementation to understand the usage of this function. The Chinese translation of Example is: show

C++ program to get the imaginary part of a given complex number C++ program to get the imaginary part of a given complex number Sep 06, 2023 pm 06:05 PM

Modern science relies heavily on the concept of plural numbers, which was first established in the early 17th century by Girolamo Cardano, who introduced it in the 16th century. The formula for complex numbers is a+ib, where a holds the html code and b is a real number. A complex number is said to have two parts: the real part <a> and the imaginary part (<ib>). The value of i or iota is √-1. The plural class in C++ is a class used to represent complex numbers. The complex class in C++ can represent and control several complex number operations. Let's take a look at how to represent and control the display of plural numbers. imag() member function As mentioned above, complex numbers are composed of real part and imaginary part. To display the real part we use real()

See all articles