-1.#Q0 shows up randomly

Question

I have the following code:

float w[n][n] = {
    {0, 0, 0.5, -1, 0, 0, 0},
    {0, 0, 1.5, -2, 0, 0, 0},
    {0, 0, 0, 0, 1, -1, 0},
    {0, 0, 0, 0, 3, -4, 0},
    {0, 0, 0, 0, 0, 0, 1},
    {0, 0, 0, 0, 0, 0, -3}
};
float x[ne] = {2, -1};
float d = 1;
float alpha = 0.1;
float in[n];
float delta[n];
float a[n];
float sum;

for(j = ne; j <= n; j++) {
    for(i = 0; i <= n; i++) {
        in[j] += w[i][j] * a[i];
    }
    a[j] = g(in[j]);
}

for(i = 0; i < n; i++) {
    printf("a[%d] = %.3f\n", i+1, a[i]);
}

delta[n-1] = d - a[n-1];

for(i = n-2; i >= ne; i--) {
    for(j = 0; j <= n; j++) {
        sum += w[i][j] * delta[j];
    }
    delta[i] = g(in[i]) * (1 - g(in[i])) * sum;
    printf("delta[%d] = %.3f\n", i+1, delta[i]);
    sum = 0;
 }

printf("\n\n");

for(i = 0; i < n; i++) {
    for(j = 0; j < n; j++) {
        if (w[i][j] != 0) {
            w[i][j] = w[i][j] + alpha * a[i] * delta[j];
            printf("w[%d][%d] = %.3f\n", i+1, j+1, w[i][j]);
        }
    }
}

I keep getting some variation of this output:

a[1] = 2.000
a[2] = -1.000
a[3] = 0.378
a[4] = 0.500
a[5] = 0.867
a[6] = 0.085
a[7] = 0.649

delta[6] = -1.#QO
delta[5] = -1.#QO
delta[4] = -1.#QO
delta[3] = -1.#QO

w[1][3] = -1.#QO
w[1][4] = -1.#QO
w[2][3] = -1.#QO
w[2][4] = -1.#QO
w[3][5] = -1.#QO
w[3][6] = -1.#QO
w[4][5] = -1.#QO
w[4][6] = -1.#QO
w[5][7] = 1.030
w[6][7] = -2.997

It shows up in different places everytime I compile, knowing that the first time it was giving me the right answer, I'm not sure what went wrong?

https://stackoverflow.com/questions/5939573/what-float-value-makes-sprintf-s-produce-1-qo — Zan Lynx, Jun 08 '17 at 04:43
@mari your comment includes the definition, yet not the initialization. Post their initialization too. — chux - Reinstate Monica, Jun 08 '17 at 04:48
@chux thanks for pointing that out, it turned out I haven't initialized `sum` which was the cause of the problem! — mari, Jun 08 '17 at 04:52
@mari Your can post your own answer and accept to to help close this post. — chux - Reinstate Monica, Jun 08 '17 at 04:54

score 2 · Answer 1 · answered Jun 08 '17 at 04:46

That code is not complete. Why did you leave out the declarations all the variables? Because they "aren't important?"

Note that those declarations ARE IMPORTANT.

I can't tell but I predict that d is not set to 0 and is picking up 0xFFFFFFFF from the stack. Which is Not A Number in floating point. A NaN value in a calculation always results in NaN results.

-1.#Q0 shows up randomly

1 Answers1