All Questions
Tagged with statistics linear-algebra
858
questions
0
votes
0
answers
10
views
Encouraging sparsity at block level or element-wise level?
I have an objective function $f(W)$, where $W$ is a $Kp \times Kp$ matrix. We can view $W$ is a $p \times p$ block matrix, where each block has the dimension $K \times K$. Now to optimize $f(W)$, I ...
0
votes
1
answer
20
views
How to sidestep of undifferentiability of Frobenius norm at 0 in the numerical analysis?
I am currently doing the l-bfgs-b optimization algorithm. I have my objective function. I also need to get the gradient of the objective function. Some part of my objective function is Frobenius norm ...
2
votes
0
answers
56
views
Covariances not below $\Sigma$
$\Sigma_0,\Sigma_1,\dots,\Sigma_K$ are real covariance matrices.
I’m interested in the set of matrices
$$\bigcap_{k=1}^K \left\{x: 0 \preceq x \preceq \Sigma_0, \ x\not\prec\Sigma_k\right\}.$$
I’m ...
1
vote
0
answers
64
views
Relation between values of $ξ_i$ and $\alpha_i$ in SVM?
I have a question in about a property of support vectors of SVM which is stated in subsection "12.2.1 Computing the Support Vector Classifier" of "The Elements of Statistical Learning&...
0
votes
0
answers
39
views
least squares minimum test error solution
assume we want to learn a model $y=x^T \beta + \varepsilon $
where
$\beta \in \mathbb{R}^d$ is constant
$ x \in \mathbb{R}^d$ is the input vector with Gaussian distribution $\mathcal{N}(0,\Sigma_x)$ ...
1
vote
0
answers
22
views
What does the spectral norm of a Wigner matrix converge to when the variances are not renormalised?
It seems that it is well known that for a $NxN$ Wigner matrix - that is a matrix that is symmetric (or Hermitian, but I am only interested in the case where all the entries are real) and has i.i.d. ...
-1
votes
1
answer
47
views
What is the derivative of the $\ell_1$ norm of the matrix? [duplicate]
The question is short. I have a square matrix $W$. I know $\|W\|_1$ means the usual $\ell_1$ norm, which means the sum of the absolute value of elements of the matrix. Now I want to compute the ...
0
votes
0
answers
23
views
What is the intuition to create an orthogonal design matrix in an iterative way (without Gram-Schmidt)?
Let $X_1, X_2, \dots, X_n$ be $n$ observations between $[-1,1]$ and
$X_i \ne X_j$ if $i\ne j$. Let $\phi_0(x)=1$, $\phi_1(x) =
> 2(x-a_1)\phi_0(x)$.
When $r\ge 1$, $\phi_{r+1}(x)=2(x-a_{r+1})\...
0
votes
0
answers
31
views
How to Upper Bound the Spectral Norm of $\left(XX^T\right)^{-1}\left(XX^T\right)^{-1}X$?
I have an observation matrix $ X \in \mathbb{R}^{n \times n}$. Considering $XX^T$, this matrix can be seen as a correlation matrix between individuals, so it generally has elements close to the ...
1
vote
1
answer
63
views
PCA Reconstruction Properties
Let $X \in \mathbb{R}^{n \times d}$ be our data matrix where $n$ is the number of examples and $d$ is the feature dimension. Applying PCA to $X$, we get a low-dimensional representation $A \in \mathbb{...
2
votes
0
answers
49
views
How to show this discrete quadratic equation converges?
So I have a discrete process
$V_{k}=AV_{k-1}A^T+C$
where $C$ is a constant and $V$ is symmetric (this is supposed to be the update for the state covariance of a discrete stochastic process). I read ...
1
vote
1
answer
45
views
If $X^TX\beta=X^TY$, then $X\beta$ is independent of $\beta$
This question is motivated by linear statistical inference, and more specifically, the normal equation for a least squares estimate and estimable functions. But it boils down to pure linear algebra.
...
1
vote
1
answer
29
views
Compute inverse of a special 2 by 2 block matrix.
Let
$$X\in\mathbb{R}^p,\quad
\tilde{X} = (1, X^{\top})^{\top}\in\mathbb{R}^{p+1},\quad \tilde{\Sigma}=\mathbb{E} \left[\tilde{X} \tilde{X}^{\top}\right]\in\mathbb{R}^{(p+1)\times (p+1)}
$$
$$
(\tilde{...
5
votes
1
answer
204
views
Rigorous Mathematical foundations of Machine Learning / Deep Learning / Neural Networks
I am an Engineering Graduate (with a strong background in Probability/Measure Theory, Linear Algebra and Calculus) wanting to dig deep into Deep Learning and Neural Networks, and I'm looking for ...
0
votes
0
answers
10
views
Given two matrices, where the average of correlations for each column is $0$, how to determine the range of average correlation values for each row?
Given two matrices A and B, where $mean(corr(A_{\cdot,i},B_{\cdot,i}))$ is equal to $0$, how to determine the range of $mean(corr(A_{j,\cdot},B_{j,\cdot}))$ ?