7
$\begingroup$

I have a system of non-linear equations

$ a_1=f_0 g_1$

$a_2=f_1 g_1 + f_0 g_2$

$a_3=f_2 g_1 + f_6 g_2 + f_0 g_3 $

$a_4=f_3 g_1 + f_7 g_2 + f_6 g_3 + f_0 g_4 $

$a_5=f_4 g_1 + f_8 g_2 + f_7 g_3 + f_6 g_4 + f_0 g_5$

$a_6=f_5 g_1 + f_9 g_2 + f_8 g_3 + f_7 g4 + f_6 g_5 + f_0 g_6$

$a_7=f_5 g_3 + f_9 g_4 + f_8 g_5 + f_2 g_6 + f_0 g_7 $

$a_8=f_5 g_5 + f_4 g_6 + f_7 g_7 + f_0 g_8$

$a_9=f_5 g_7 + f_8 g_8 + f_0 g_9$

$a_{10}=f_0 g_{10} + f_5 g_9$

here both $f=(f_0,..,f_5,,..f_9)$ and $g=(g_1,…,g_{10})$ are unknown.

My problem is to figure out when this system has no solution, a unique or multiple solutions.

What is known is that this system of equations allows to decompose itself as

$\left(\begin{array}{c} a_1\\ a_2 \\ \vdots \\a_9 \\a_{10} \end{array} \right) = \begin{bmatrix} f_0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 \\ f_1 & f_0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 \\ f_2 & f_6& f_0& 0& 0& 0& 0& 0 &0 & 0 \\ f_3 & f_7 & f_6 & f_0 & 0 & 0 & 0 & 0 & 0& 0 \\ f_4 & f_8 & f_7 & f_6 & f_0& 0& 0& 0& 0& 0 \\ f_5 & f_9 & f_8 & f_7 & f_6& f_0& 0& 0& 0& 0 \\ 0 & 0 & f_5 & f_9 & f_8 & f_2& f_0& 0 & 0& 0 \\ 0 & 0 & 0 & 0 & f_5 & f_4 & f_7 & f_0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 & f_5& f_8 & f_0 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & f_5 & f_0 \end{bmatrix} \times \left( \begin{array}{c} g_1 \\ g_2 \\ \vdots \\ g_9 \\ g_{10} \end{array} \right) $

Then whenever I take only vector of $f$ as unknowns (and so $g$ are the parameters), or vice versa, there is a unique solution for a generic vector $a$, similarly to how it works for linear systems of equations. While when I am letting one extra element of $f$ or of $g$ be "a parameter", i.e. there are more equations than there are unknown $f$ or unknown $g$, there is no solution.

But I have no idea what happens when both $f$ and $g$ are unknown. I have then 10 equations and 20 unknowns, but it is not clear to me that it automatically should lead to a continuum of solutions. (And I cannot solve this system explicitly..)

If there is a continuum of solutions (for a "generic" $a$), it seems to be an interesting case: a system of $N$ non-linear equations and $M(>N)$ unknowns does not have a solution (for generic vector $a$) whenever the size of the smallest of two sets of unknowns is less than $N$. Or it has a continuum of solutions in the complementary case, but there is no case in-between -- with a unique solution.

Are there any known results for such system of non-linear equations that are "decomposable"?
I would appreciate any opinions on this!

$\endgroup$
  • $\begingroup$ This doesn't exploit the "decomposable" structure, but have you looked into Gröbner bases for your system of 10 quadratic equations? en.wikipedia.org/wiki/Gr%C3%B6bner_basis $\endgroup$ – j.c. Sep 22 '15 at 16:25
  • 4
    $\begingroup$ where do this equations come from? If you have a solution you can multiply all f's with a constant and all g's with the inverse of the constant to get a new solution. $\endgroup$ – user35593 Sep 22 '15 at 19:11
  • $\begingroup$ Thank you j.c. for editing, I was trying to find the Gröbner basis, but was unsuccessful (in fact even doing this in Mathematica did not help), so I started to look for more general ideas... $\endgroup$ – Kass Sep 22 '15 at 19:28
  • $\begingroup$ @user35593: My bad! you are perfectly right, that one can generate easily a continuum of solutions, provided we have at least one. In fact there is less freedom than what I stated. My $g$ sum up to 1. And $f$ belong to [0,1]. As I have been loosing hope to get a solution that satisfies these extra constraints, I started to get rid of them and see for unconstrained solutions. I guess I should edit my question. $\endgroup$ – Kass Sep 22 '15 at 19:31
  • $\begingroup$ So the system has the form $B(f,g)=a$ for a bilinear map $$B:\mathbb{R}^{10}\times \mathbb{R}^{10} \to \mathbb{R}^{10}$$ and $B(f,\cdot)$ is invertible iff $f_0\neq0$. $\endgroup$ – Pietro Majer Dec 4 '15 at 8:37
8
+50
$\begingroup$

The linear-in-$g$ system is uniquely solvable whenever $f_0\ne0$. therefore the solution set is parametrized by $f$ in the form $$g=f_0^{-10}P(f;a),$$ where $P$ is linear in $a$, polynomial in $f$ and homogenenous of degree $9$. Not only it is a continuum, but it has dimension $10$.

When $f_0=0$, there is no solution in general, say for $a_1\ne0$.

$\endgroup$
  • $\begingroup$ So the solution set is a graph over $\mathbb{R}^{10}$, plus (if $a_1=0$) a 9-dimensional vertical component corresponding to $f_0=0$ and $f_1 f_2 f_5 f_6f_7f_8\neq0$; plus possibly other smaller components according to the form of $a$. $\endgroup$ – Pietro Majer Dec 4 '15 at 8:50
1
$\begingroup$

How about setting $f_0=1$ and $f_1=t$ where $t$ is a parameter and the rest of the $f_i=0?$ Then $g_i=a_i$ for all $i $ except $g_2=a_2-ta_1.$ That seems like a continuum of solutions, even projectively (i.e. up to rescaling.)

It might be interesting to ask if there are always $9$ degrees of freedom (or $10$ with rescaling allowed.)

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.