Inverse of a matrix fails

[danielmartensson]

Inverse of a matrix fails

[DanielMartensson]

Can you write out what the value is at this line

CControl/src/CControl/Sources/LinearAlgebra/lup.c

Line 40 in 5510477

if (fabsf(LU[row * P[i] + i]) < FLT_EPSILON)

When lup.c file returns 0, then the inv.c returns 0 as well.

Edit:

I did a quick check. I found the issue

/*
 ============================================================================
 Name        : inv.c
 Author      : Daniel Mårtensson
 Version     : 1.0
 Copyright   : MIT
 Description : Turn A into inverse A^(-1)
 ============================================================================
 */

#include "CControl/Headers/Functions.h"

int main() {

	// Matrix A
	float A[2*2] = {1, 1, 1, 1};

	clock_t start, end;
	float cpu_time_used;
	start = clock();
	int status = inv(A, 2);
	printf("Status on matrix A(1 == success, 0 == fail): %d\n", status);
	end = clock();
	cpu_time_used = ((float) (end - start)) / CLOCKS_PER_SEC;
	printf("\nTotal speed  was %f\n", cpu_time_used);

	// Print A
	printf("A^(-1)\n");
	print(A, 2, 2);

	return EXIT_SUCCESS;
}

Output:

Status on matrix A(1 == success, 0 == fail): 1

Total speed  was 0.000000
A^(-1)
1.#INF00000000000000	-1.#INF00000000000000	
-1.#INF00000000000000	1.#INF00000000000000	

The issue is here:
Question: Can you be so kind to do an if-statement and check if we are dividing with 0 ? If we are dividing with zero, then the static function should return uint8_t or bool as 0 or false. As a check only. Can you do this? Send me then a pull request?

CControl/src/CControl/Sources/LinearAlgebra/inv.c

Line 68 in 5510477

x[i] = x[i] / LU[row * P[i] + i];

Edit:

I have fixed the issue.
Please re-open the issue if the issue still occurs.

[ooaj]

Aha. Great! Thanks @DanielMartensson. Works perfectly now.

[DanielMartensson]

Aha. Great! Thanks @DanielMartensson. Works perfectly now.

Thank you @ooaj for finding this issue.
If you want, you could help me implement a Kernel Density Estimation algorithm. It's very easy.

Here is the formula:

K is the gaussian propability density function.

https://en.wikipedia.org/wiki/Kernel_density_estimation#Definition

The theory: See between 0:0 to 2:00 https://www.youtube.com/watch?v=DCgPRaIDYXA&t=1s&ab_channel=KimberlyFessel

Assume that you have a mesurement float my_measurement[100] and you want to find the correct value, because it's a nonlinear measurement. If you plot float my_measurement[100] you can see two gaussian curves, but which one should you select? Just use KDE algorithm as above.

Can you do that?

[ooaj]

Sure. I can try to implement it. I have a very busy weekend, so if that's not a problem I can start working on it on Monday. I imagine you are going to be using it for your Particle Filter, right?