Mathematics

Associated with Math.SE; for both general discussion & math qu...
Oct 4, 2021 06:05
This is the book: book-wright-ma.github.io/Book-WM-20210422.pdf, page 89 - 91
Oct 4, 2021 06:02
In the textbook I am reading right now, it says:

(Restricted Strong Convexity). The matrix A satisfies the restricted strong convexity (RSC) condition of order $k$, with parameters $\mu>0$, $\alpha \geq 1$, if for every $\mathrm{I}$ of size at most $k$ and for all nonzero $\boldsymbol{h}$ satisfying $\left\|\boldsymbol{h}_{\mathbf{| c}}\right\|_{I} \leq$ $\alpha\left\|\boldsymbol{h}_{I}\right\|_{1}$
$$
\|\boldsymbol{A} \boldsymbol{h}\|_{2}^{2} \geq \mu\|\boldsymbol{h}\|_{2}^{2}
$$

However this is said to satisfy the Nullspace property:
Oct 4, 2021 02:54
$\left\|\left(\boldsymbol{X}^{T} \boldsymbol{X}\right)^{-1} \boldsymbol{X}^{T} \boldsymbol{v}\right\|_{2} \leq \frac{1}{\sigma_{\min }(\boldsymbol{X})}\|\boldsymbol{v}\|_{2}$ Does anyone have any hints as to how I can proceed in proving this inequality?
Mar 4, 2021 14:51
Where there is an answer in the comments, but the answer says that there is a nice general explanation. So that got me wondering, does the minimum norm imply a better fit for a regression problem?
Mar 4, 2021 14:50
Mar 4, 2021 14:49
Is it then a worse regression ?
Mar 4, 2021 14:48
But what if you minimize $||Ax-b||^2$ and you obtain an $x$ solution with a higher norm
Mar 4, 2021 14:48
Let us say $Ax = b$ , where $A$ is rank-deficient
Mar 4, 2021 14:47
Why does one want the solution to the minimum norm solution?
Mar 4, 2021 14:46
Given a least squares regression problem
Mar 4, 2021 14:46
Yeah
Mar 4, 2021 14:45
Hey, I was just wondering. In Linear Algebra, why does one want minimum norm solution ?
Oct 18, 2020 19:26
Thank you!
Oct 18, 2020 19:26
Indeed!
Oct 18, 2020 19:20
Any hints hehe?
Oct 18, 2020 19:18
@skillpatrol With epsilon-delta?
Oct 18, 2020 19:15
But I am still a little lost on how I can show that my limit goes to infinity
Oct 18, 2020 19:14
Thanks for the explanation
Oct 18, 2020 19:14
Indeed
Oct 18, 2020 19:12
Let us say $\frac{1}{0} = x$ but $1 \ne 0*x$
Oct 18, 2020 19:10
That is possible, but I don't really know the correct answer to why that is allowed and dividing by zero is not allowed
Oct 18, 2020 19:08
Because there is nothing to divide on
Oct 18, 2020 19:08
It just doesn't make sense?
Oct 18, 2020 19:06
It means that the function is undefined at that point?
Oct 18, 2020 19:05
That is a good point
Oct 18, 2020 19:04
If I insert $a$ for $x$ I will get $\frac{1}{0}$ which is not possible ( I think )
Oct 18, 2020 19:01
That I do
Oct 18, 2020 18:58
Which means that the limit I have got will give a division by zero, which is not allowed
Oct 18, 2020 18:57
if x approaches a
Oct 18, 2020 18:57
Wouldn't it be zero?
Oct 18, 2020 18:55
Really lost on this one, sorry
Oct 18, 2020 18:54
Not defined?
Oct 18, 2020 18:51
I was thinking L'Hopital but that wouldn't get me very far
Oct 18, 2020 18:49
Found a task in my Professors exercises which I am not sure I understand:
Find $lim_{x \rightarrow a+0} \frac{1}{(x-a)^{2k}} \, \, \, k \in \mathbb{N}$
How would one proceed with such a task?
Oct 1, 2020 10:54
math.stackexchange.com/questions/1728376/… Trying to solve exactly the same problem as this one. I defined $\delta = \frac{1}{2}$, which means that $|x| < 1/2$. And $|x^2 - 1| > \frac{3}{4}$. But how does one proceed from this to create the relationship between $\delta$ and $\epsilon$?
Sep 29, 2020 21:15
Sorry, if I am asking very obvious questions
Sep 29, 2020 21:14
I think I might just be stupid, would this mean that $s$ could have been $\frac{a}{a^2-2b^2}$ and $t$ could have been $\frac{b}{a^2-2b^2}$. And as long as $s,t$ were still rational numbers, they would be in $S$?
Sep 29, 2020 21:00
I see! And then because of how S is defined it would $s-t\sqrt(2)$ part will show that it is in S?
Sep 29, 2020 20:49
Hello everyone! Suppose I have a set defined by $S= {s + t\sqrt{2} \, \, | s,t \in \mathbb{Q}}$ And $x \neq 0, x \in S$, how can I show that $\frac{1}{x} \in S$ as well? Any guideline or tips would be very much appreciated.
May 19, 2020 14:11
Can type it in MathJax; $x- 2y + 3z =1$
May 19, 2020 14:08
Hey! Quick question is x-2y+3z=1 a subspace of R^3?
 

 Calvin's Chat Room

A place to sometimes discuss mathematics, fluids, mathematical...
Oct 1, 2020 13:04
Still stuck, can’t seem to wrap my head around it
Oct 1, 2020 11:54
Any further advice? Sorry for the stupidity
Oct 1, 2020 11:53
If I I set $4x^2/3 \le \epsilon$ I get $x \le \sqrt{3 \epsilon} / 2$
Oct 1, 2020 11:52
Hmm
Oct 1, 2020 11:25
Sorry if I am asking very obvious questions, I am having a hard time understanding Andre's suggestion
Oct 1, 2020 11:25
I tried making sense of it so, $|x^2-1| > 3/4$, which means that $\frac{x^2}{x^2-1} \lt \frac{4x^2}{3}$ which in turn means that $|f(x) - L| \le \frac{4x^2}{3}$. But how do I go forward from here?
Oct 1, 2020 11:20
However, I am struggling to proceed from this. Do you have any advice?
Oct 1, 2020 11:19
Ah, that makes sense. To limit $\delta$
Oct 1, 2020 11:17
https://math.stackexchange.com/questions/1728376/proof-limit-of-frac1x2-1-1-using-epsilon-delta-as-x-0-t Trying to solve exactly the same problem as this one. I defined $\delta = \frac{1}{2}$, which means that $|x| < 1/2$. And $|x^2 - 1| > \frac{3}{4}$. But how does one proceed from this to create the relationship between $\delta$ and $\epsilon$?