assignmentutor™ 为您的留学生涯保驾护航 在代写信息论information theory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。我们的专家在代写信息论information theory代写方面经验极为丰富，各种代写信息论information theory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
assignmentutor™您的专属作业导师

## 数学代写|信息论代写information theory代考|Examples of application of general methods for computation of channel capacity

Example 3.1. For simplicity let there be only two symbols initially: $m=2 ; y=1,2$, which correspond to different costs
$$c(1)=b-a, \quad c(2)=b+a \text {. }$$
In this case, the partition function (3.3.11) is equal to
$$Z=e^{-\beta b+\beta a}+e^{-\beta b-\beta a}=2 e^{-\beta b} \cosh \beta a .$$
By formula (3.3.12) the free energy
$$F=b-T \ln 2-T \ln \cosh \frac{a}{T}$$
corresponds to it. Applying formulae (3.3.12), (3.3.13) we find entropy and average energy
\begin{aligned} &C=H_{y}(T)=\ln 2+\ln \cosh \frac{a}{T}-\frac{a}{T} \tanh \frac{a}{T} \ &R(T)=b-a \tanh \frac{a}{T} \end{aligned}
The graphs of these functions are presented on Figure 3.1. It is also shown there how to determine channel capacity with the given level of $\operatorname{costs} R=R_{0}$ in a simple graphical way.

When temperature changes from 0 to $\infty$, entropy changes from 0 to $\ln 2$ and energy goes from $b-a$ to $b$ (where $a>0$ ). Also in a more general case as $T \rightarrow \infty$ entropy $H_{y}$ goes to limit $\ln m$. This value appears to be the channel capacity corresponding to the absence of constraints. Also, this value cannot be exceeded for any average cost.

The additive parameter $b$ implicated in the cost function (3.4.1) is not essential. Neither entropy nor a probability distribution depends on it. This corresponds to the fact that the additive constant from the expression for energy (recall that $R$ is an analog of energy) can be chosen arbitrarily. In the example in question the optimal probability distribution has the form
$$P(1,2)=e^{\pm a / T} / 2 \cosh (a / T)$$
according to (3.3.14). The determined functions (3.4.2) can be used in cases, for which there exists sequence $y^{L}=\left(y_{1}, \ldots, y_{L}\right)$ of length $L$ that consists of symbols described above. If the number of distinct elementary symbols equals 2 , then the number of distinct sequences is evidently equal to $m=2^{L}$. Next we assume that the costs imposed on an entire sequence are the sum of the costs imposed on symbols, which constitute this sequence. Hence,
$$c\left(y^{L}\right)=\sum_{i=1}^{L} c\left(y_{i}\right)$$

## 数学代写|信息论代写information theory代考|Methods of potentials in the case of a large number of parameters

The method of potentials worded in Sections $3.3$ and $3.4$ can be generalized to more difficult cases when there are a larger number of external parameters, i.e., this method resembles methods usually used in statistical thermodynamics even more.

Here we outline possible ways to realize the above generalization and postpone a more elaborated analysis to the next chapter.

Let cost function $c(y)=c(y, a)$ depend on a numeric parameter $a$ now and be differentiable with respect to this parameter. Then free energy
$$F(T, a)=-T \ln \sum_{y} e^{-c(y, u) / T}$$
and other functions introduced in Section $3.4$ will be dependent not only on temperature $T$ (or parameter $\beta=1 / T$ ), but also on the value of $a$. The same holds true for the optimal distribution (3.3.14) now having the form
$$P(y \mid a)=\exp \left[\frac{F(T, a)-c(y, a)}{T}\right] .$$
Certainly, formula (3.3.15) and other results from Section $3.3$ will remain valid if we account parameter $a$ to be constant when varying parameter $T$, i.e. if regular derivatives are replaced with partial ones.
Hence, entropy of distribution (3.5.2) is equal to
$$H_{y}=-\frac{\partial F(T, a)}{\partial T}$$
Now in addition to these results we can derive a formula of partial derivatives taken with respect to the new parameter $a$.

## 数学代写|信息论代写information theory代考|Capacity of a noiseless channel with penalties

1. In Sections $3.2$ and $3.3$ we considered capacity of a discrete channel without noise, but with penalties. As it was stated, computation of that channel capacity can be reduced to solving the first variational problem of information theory. The results derived here can be generalized to the case of arbitrary random variables, which can assume continuous values, in particular. Some formulae relative to the generalized version of the formula and provided in Section $1.6$ give us some hints how to do so.
We suppose that there is given a noiseless channel (not necessarily discrete) if there are given measurable space $(X, F)$, random variable $\xi$ taking values from the indicated space, $F$-measurable function $c(\xi)$ called a cost function, and also measure $v$ on $(X, F)$ (normalization of $v$ is not required). We define channel capacity $C(a)$ for the level of losses $a$ as the maximum value of entropy (1.6.9):
$$H_{\xi}=-\int P(d \xi) \ln \frac{P(d \xi)}{v(d \xi)},$$
compatible with the constraint
$$\int c(\xi) P(d \xi)=a .$$
Basically, the given variational problem is solved in the same way as it was done in Section 3.3. But note that partial derivatives are replaced with variational derivatives in this modified approach. After variational differentiation with respect to $P(d x)$ instead of (3.3.4) we will have the extremum condition:
$$\ln \frac{P(d \xi)}{v(d \xi)}=\beta F-\beta c(\xi)$$
where $\beta F=-1-\gamma$.
From here we obtain the extremum distribution
$$P(d \xi)=e^{\beta F-\beta c(\xi)} v(d \xi) .$$
Averaging (3.6.3) and taking into account (3.6.1), (3.6.2) we obtain that
$$H_{\xi}=\beta \mathbb{E}[c(\xi)]-\beta F, \quad C=\beta R-\beta F .$$

## 数学代写|信息论代写information theory代考|Examples of application of general methods for computation of channel capacity

$$c(1)=b-a, \quad c(2)=b+a .$$

$$Z=e^{-\beta b+\beta a}+e^{-\beta b-\beta a}=2 e^{-\beta b} \cosh \beta a$$

$$F=b-T \ln 2-T \ln \cosh \frac{a}{T}$$

$$C=H_{y}(T)=\ln 2+\ln \cosh \frac{a}{T}-\frac{a}{T} \tanh \frac{a}{T} \quad R(T)=b-a \tanh \frac{a}{T}$$

$$P(1,2)=e^{\pm a / T} / 2 \cosh (a / T)$$

$$c\left(y^{L}\right)=\sum_{i=1}^{L} c\left(y_{i}\right)$$

## 数学代写|信息论代写information theory代考|Methods of potentials in the case of a large number of parameters

$$F(T, a)=-T \ln \sum_{y} e^{-c(y, u) / T}$$

$$P(y \mid a)=\exp \left[\frac{F(T, a)-c(y, a)}{T}\right]$$

$$H_{y}=-\frac{\partial F(T, a)}{\partial T}$$

## 数学代写|信息论代写information theory代考|Capacity of a noiseless channel with penalties

1. 在部分 $3.2$ 和 $3.3$ 我们考虑了没有㖏声但有惩罚的离散信道的容量。如前所述，信道容量的计算可以简化为解决信息论的第一变分问 题。这里得出的结果可以推广到任意随机变量的情况，特别是可以假设连续值。一些与公式的广义版本相关的公式，并在部分提供 $1.6$ 给我们一些提示。
2. 如果给定可测量的空间，我们假设给定一个无樔声通道 (不一定是离散的) $(X, F)$ ，随机变量 $\xi$ 从指定的空间取值， $F$ – 可测量函数 $c(\xi)$ 称为成本函数，也可以测量 $v$ 上 $(X, F)$ (归一化 $v$ 不需要) 。我们定义渠道容量 $C(a)$ 对于损失水平 $a$ 作为樀的最大值 (1.6.9) :
3. $$4. H_{\xi}=-\int P(d \xi) \ln \frac{P(d \xi)}{v(d \xi)}, 5.$$
6. 与约束兼容
7. $$8. \int c(\xi) P(d \xi)=a 9.$$
10. 基本上，给定的变分问题的解决方法与第 $3.3$ 节中的方法相同。但请注意，在这种修改后的方法中，偏导数被变分导数取代。在变 分微分之后 $P(d x)$ 代替 (3.3.4) 我们将有极值条件:
11. $$12. \ln \frac{P(d \xi)}{v(d \xi)}=\beta F-\beta c(\xi) 13.$$
14. 在哪里 $\beta F=-1-\gamma$.
15. 从这里我们得到极值分布
16. $$17. P(d \xi)=e^{\beta F-\beta c(\xi)} v(d \xi) 18.$$
19. 平均 (3.6.3) 并考虑 (3.6.1)， (3.6.2) 我们得到
20. $$21. H_{\xi}=\beta \mathbb{E}[c(\xi)]-\beta F, \quad C=\beta R-\beta F 22.$$

## 有限元方法代写

assignmentutor™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。