统计代写|主成分分析代写Principal Component Analysis代考|ENVX2001

statistics-lab™ 为您的留学生涯保驾护航 在代写主成分分析Principal Component Analysis方面已经树立了自己的口碑, 保证靠谱, 高质且原创的统计Statistics代写服务。我们的专家在代写主成分分析Principal Component Analysis代写方面经验极为丰富，各种代写主成分分析Principal Component Analysis相关的作业也就用不着说。

• Statistical Inference 统计推断
• Statistical Computing 统计计算
• (Generalized) Linear Models 广义线性模型
• Statistical Machine Learning 统计机器学习
• Longitudinal Data Analysis 纵向数据分析
• Foundations of Data Science 数据科学基础

统计代写|主成分分析代写Principal Component Analysis代考|Explained Variance

Finally, we introduce bounds for the explained variance $E V(X)$. Two results are obtained. The first is general and applicable to any basis $X$, not limited to sparse ones. The second is tailored to SPCArt.

Theorem 12 Let rank-r SVD of $A \in \mathbb{R}^{n \times p}$ be $U \Sigma V^{T}, \Sigma \in \mathbb{R}^{r \times r}$. Given $X \in \mathbb{R}^{p \times r}$, assume the $S V D$ of $X^{T} V$ to be $W D Q^{T}, D \in \mathbb{R}^{r \times r}, d_{\min }=\min {i} D{i i}$, then
$$d_{\min }^{2} \cdot E V(V) \leq E V(X),$$
and $E V(V)=\sum_{i} \Sigma_{i i^{\circ}}^{2}$
The theorem can be interpreted as follows. If $X$ is a basis that approximates the rotated PCA loadings well, then $d_{\min }$ will be close to one, and so the variance explained by $X$ is close to that explained by PCA. Note that the variance explained by PCA loadings is the largest value that is possible to be achieved by an orthonormal basis. Conversely, if $X$ deviates greatly from the rotated PCA loadings, then $d_{\min }$ tends to zero, so the variance explained by $X$ is not guaranteed to be large. Thus, the less the sparse loadings deviate from the rotated PCA loadings, the more variance they explain.

When SPCArt converges, i.e., $X_{i}=T_{\lambda}\left(Z_{i}\right) /\left|T_{\lambda}\left(Z_{i}\right)\right|_{2}$, where $Z=V R^{T}$, and $R=\operatorname{Polar}\left(X^{T} V\right)$ hold simultaneously, there is another estimation (mainly valid for $\mathrm{T}$-en).

Theorem 13 Let $C=Z^{T} X$, i.e., $C_{i j}=\cos \left(\theta\left(Z_{i}, X_{j}\right)\right)$, and let $\bar{C}$ be the diagonalremoved version. Assume $\forall i, \theta\left(Z_{i}, X_{i}\right)=\theta$ and $\sum_{j}^{r} C_{i j}^{2} \leq 1$, then
$$\left(\cos ^{2}(\theta)-\sqrt{r-1} \sin (2 \theta)\right) \cdot E V(V) \leq E V(X) .$$
When $\theta$ is sufficiently small,
$$\left(\cos ^{2}(\theta)-O(\theta)\right) \cdot E V(V) \leq E V(X) .$$
Since the sparse loadings are obtained by truncating small entries of the rotated PCA loadings, and $\theta$ is the deviation angle, the theorem implies that if the deviation

is small then the explained variance is close to that of $\mathrm{PCA}$, as $\cos ^{2}(\theta) \approx 1$. For example, if the truncated energy $|\bar{z}|_{2}^{2}=\sin ^{2}(\theta)$ is approximately $0.05$, then $95 \%$ $\mathrm{EV}$ of $\mathrm{PCA}$ loadings is guaranteed.

The assumptions $\theta\left(Z_{i}, X_{i}\right)=\theta$ and $\sum_{j}^{r} C_{i j}^{2} \leq 1, \forall i$, are broadly satisfied by T-en using small $\lambda$. Uniform deviation $\theta\left(Z_{i}, X_{i}\right)=\theta \forall i$ can be achieved by T-en, as indicated by Proposition 11. $\sum_{j}^{r} C_{i j}^{2} \leq 1$ means the sum of projected length is less than 1 when $Z_{i}$ is projected onto each $X_{j}$. It is satisfied if $X$ is exactly orthogonal, whereas it is likely satisfied if $X$ is nearly orthogonal (note $Z_{i}$ may not lie in the subspace spanned by $X$ ), which can be achieved by setting small $\lambda$ according to Proposition 6. In this case, about $(1-\lambda) E V(V)$ is guaranteed.

In practice, we prefer CPEV [21] to EV. CPEV measures the variance explained by subspace rather than basis. Since it is also the projected length of $A$ onto the subspace spanned by $X$, the higher CPEV, the better $X$ represents the data. If $X$ is not an orthogonal basis, EV may overestimate or underestimate the variance. However, if $X$ is nearly orthogonal, the difference is small, and it is nearly proportional to CPEV.

统计代写|主成分分析代写Principal Component Analysis代考|A Unified View to Some Prior Work

A series of methods: PCA [10], SCoTLASS [11], SPCA [29], GPower [13], $\mathrm{rSVD}[21]$, TPower [25], SPC [24], and SPCArt, although proposed independently and formulated in various forms, can be derived from the common source of Theorem 1, the Eckart-Young Theorem. Most of them can be seen as the problems of matrix approximation (1), with different sparsity penalties. Most of them have two matrix variables, and the solutions of them can usually be obtained by an alternating scheme: fixing one and solving the other. Similar to SPCArt, the two subproblems are a sparsity penalized/constrained regression problem and a Procrustes problem.
PCA [10]. Since $Y^{}=A X^{}$, substituting $Y=A X$ into (1) and optimizing $X$, the problem is equivalent to
$$\max {X} \operatorname{tr}\left(X^{T} A^{T} A X\right), \text { s.t. } X^{T} X=I .$$ By the Ky Fan theorem [7], $X^{}=V{1: r} R, \forall R^{T} R=I$. If $A$ is a mean-removed data matrix, the special solution $X^{}=V_{1 r}$ contains exactly the $r$ loadings obtained by PCA.
SCoTLASS [11]. Constraining $X$ to be sparse in (19), we get SCotLASS
$$\max {X} \operatorname{tr}\left(X^{T} A^{T} A X\right), \text { s.t. } X^{T} X=I, \forall i,\left|X{i}\right|_{1} \leq \lambda$$
Unfortunately, this problem is not easy to solve.

统计代写|主成分分析代写Principal Component Analysis代考|Principal Component Analysis (PCA) Based

Principal component analysis (PCA) $[10,12,27]$ is an orthogonal basis transformation with the advantage that the first few principal components preserve most of the variance of the data set. This method [27], initially, calculates the covariance matrix of the given data set, and then finds the eigenvalues and eigenvectors of this matrix. Next it selects a few eigenvectors whose eigenvalues are more to form the transformation matrix to reduce the dimensions of the data set.

Suppose, there are $D$ number of band images. So, a pixel has $D$ number of different responses over different wavelengths. As a consequences, a pixel may be treated as a pattern of $D$ attributes. The main target is to reduce the dimensionality from $D$ to $d(d \ll D)$ of hyperspectral image pixel.

Let, there be a set of pattern $x_{i}$, where $x_{i} \in \mathfrak{R}^{D}, i=1,2, \ldots, N$. Assume that the data are centered, i.e., $x_{i} \Longleftarrow x_{i}-E\left{x_{i}\right}$. Conventional PCA formulates the eigenvalue problem by
$$\lambda V=\Sigma_{x} V$$
where $\lambda$ is eigenvalue, $V$ is eigenvector, $\Sigma_{x}$ is the corresponding covariance matrix over data set $x$ which is calculated by the following equation
$$\Sigma_{x}=\frac{1}{N} \sum_{i=1}^{N} x_{i} x_{i}^{T}$$
The projection on the eigenvector $V^{k}$ is calculated as
$$x_{p c}^{k}=V^{k} . x .$$
The principal component based transformation is defined as
$$y_{i}=W^{T} x_{i}$$
where $W$ is the matrix of first $d$ normalized eigenvectors of highest eigenvalues of the image covariance matrix $\Sigma_{x} . T$ denotes the transpose operation.

Here, a pattern $x_{i}$ from original $D$-dimensional space is transformed into $y_{i}$, a pattern in reduced $d$-dimensional space by choosing only the first $d$ components (eigenvectors of highest $d$ eigenvalues).

The transformed data set has two main properties which are significant to the application here. The variance in the original data set has been rearranged and reordered so that first few components contain almost all of the variance in the original data, and the components in the new feature space are uncorrelated in nature [20].

统计代写|主成分分析代写Principal Component Analysis代考|Explained Variance

(因2⁡(θ)−r−1罪⁡(2θ))⋅和在(在)≤和在(X).

(因2⁡(θ)−○(θ))⋅和在(在)≤和在(X).

统计代写|主成分分析代写Principal Component Analysis代考|Principal Component Analysis (PCA) Based

λ在=ΣX在

ΣX=1ñ∑一世=1ñX一世X一世吨

XpCķ=在ķ.X.

有限元方法代写

tatistics-lab作为专业的留学生服务机构，多年来已为美国、英国、加拿大、澳洲等留学热门地的学生提供专业的学术服务，包括但不限于Essay代写，Assignment代写，Dissertation代写，Report代写，小组作业代写，Proposal代写，Paper代写，Presentation代写，计算机作业代写，论文修改和润色，网课代做，exam代考等等。写作范围涵盖高中，本科，研究生等海外留学全阶段，辐射金融，经济学，会计学，审计学，管理学等全球99%专业科目。写作团队既有专业英语母语作者，也有海外名校硕博留学生，每位写作老师都拥有过硬的语言能力，专业的学科背景和学术写作经验。我们承诺100%原创，100%专业，100%准时，100%满意。

MATLAB代写

MATLAB 是一种用于技术计算的高性能语言。它将计算、可视化和编程集成在一个易于使用的环境中，其中问题和解决方案以熟悉的数学符号表示。典型用途包括：数学和计算算法开发建模、仿真和原型制作数据分析、探索和可视化科学和工程图形应用程序开发，包括图形用户界面构建MATLAB 是一个交互式系统，其基本数据元素是一个不需要维度的数组。这使您可以解决许多技术计算问题，尤其是那些具有矩阵和向量公式的问题，而只需用 C 或 Fortran 等标量非交互式语言编写程序所需的时间的一小部分。MATLAB 名称代表矩阵实验室。MATLAB 最初的编写目的是提供对由 LINPACK 和 EISPACK 项目开发的矩阵软件的轻松访问，这两个项目共同代表了矩阵计算软件的最新技术。MATLAB 经过多年的发展，得到了许多用户的投入。在大学环境中，它是数学、工程和科学入门和高级课程的标准教学工具。在工业领域，MATLAB 是高效研究、开发和分析的首选工具。MATLAB 具有一系列称为工具箱的特定于应用程序的解决方案。对于大多数 MATLAB 用户来说非常重要，工具箱允许您学习应用专业技术。工具箱是 MATLAB 函数（M 文件）的综合集合，可扩展 MATLAB 环境以解决特定类别的问题。可用工具箱的领域包括信号处理、控制系统、神经网络、模糊逻辑、小波、仿真等。