Find householder matrices $H_1,H_2,\cdots,H_n$ such that

$$ H_n\cdots H_1 A = L $$

where $A$ : $n \times n$ matrix and $L$ : $n \times n$ lower triangular matrix.


By defining $v_k:= [\cdots, sgn(x_k) |x_k|, \cdots]$ and $H_k := I - 2v_kv_k^T/v_k^Tv_k$, we can make

$$ H_n\cdots H_1 A = U $$

where $U$ : $n \times n$ upper triangular matrix

But I'm currently stuck at how to define $v_k$ to make RHS lower triangular.


Start from the last column, you are then in almost the same setting as for the $QR$ with upper $R$: you start with the column that will be transformed to a column with only zeros except for one coefficient (but it will be $\lambda e_n$ instead of $\lambda e_1$), and continue as usual.

Alternately, you can do that with the same algorithm as usual ($QR$ with upper $R$), by first reversing the order of the columns and rows of $A$, and doing the same afterwards on $Q$ and $R$. You can apply this to all Householder reflections, and you can check that a "reversed" reflection $H_k$ is still a Householder reflection with reversed $v_k$.

Note that if $f$ is the function that reverses the order of rows and columns, then $f(AB)=f(A)f(B)$. I don't know if it has a specific notation.


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.