### 统计代写|机器学习作业代写machine learning代考|Ambitions and Goals of Machine

statistics-lab™ 为您的留学生涯保驾护航 在代写机器学习machine learning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。我们的专家在代写机器学习machine learning方面经验极为丰富，各种代写机器学习machine learning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|机器学习作业代写machine learning代考|Training Sets and Classifiers

Let us first characterize the problem and introduce certain fundamental concepts that will keep us company us throughout the rest of the book.

Pre-Classified Training Examples Figure 1.1 shows six pies that Johnny likes, and six that he does not. In the sequel, we will refer to them as the positive and negative examples of the underlying concept. Together, they constitute a training set from which the machine is to induce a classifier-an algorithm capable of categorizing any future pie into one of the two classes: positive and negative.

The number of classes can of course be greater than just two. Thus a classifier that decides whether a landscape snapshot was taken in spring, summer, fall, or winter distinguishes four classes. Software that identifies characters scribbled

on a tablet needs at least 36 classes: 26 for letters and 10 for digits. And documentcategorization systems are capable of identifying hundreds, even thousands of different topics. The only motivation for illustrating the input to machine learning by a two-class domain was its simplicity.

Attribute Vectors To be able to communicate the training examples to the machine, we have to describe them. The most common mechanism relies on the so-called attributes. In the “pies” domain, five may be suggested: shape (circle, triangle, and square), crust-size (thin or thick), crust-shade (white, gray, or dark), filling-size (thin or thick), and filling-shade (white, gray, or dark). Table $1.1$ specifies the values of these attributes for the twelve examples in Fig. 1.1. For instance, the pie in the upper-left corner of the picture (the table calls it ex1) is described by the following conjunction:

## 统计代写|机器学习作业代写machine learning代考|Expected Benefits of the Induced Classifier

So far, we have measured the error rate by comparing the training examples’ known classes with those recommended by the classifier. Practically speaking, though, our goal is not to reclassify objects whose classes we already know; what we really want is to label future examples of whose classes we are as yet ignorant. The classifier’s anticipated performance on these is estimated experimentally. It is important to know how.

Independent Testing Examples The simplest scenario will divide the available pre-classified examples into two parts: the training set, from which the classifier is induced, and the testing set, on which it is evaluated (Fig. 1.2). Thus in the “pies” domain, with its 12 pre-classified examples, the induction may be carried out on randomly selected eight, and the testing on the remaining four. If the classifier then

“guesses” correctly the class of three testing examples (while going wrong on a single one), its performance is estimated as $75 \%$.

Reasonable though this approach may appear, it suffers from a major drawback: a random choice of eight training examples may not be sufficiently representative of the underlying concept-and the same applies to the even smaller testing set. If we induce the meaning of a mammal from a training set consisting of a whale, a dolphin, and a platypus, the learner may be led to believe that mammals live in the sea (whale, dolphin), and sometimes lay eggs (platypus), hardly an opinion a biologist will endorse. And yet, another choice of training examples may result in a classifier satisfying the highest standards. The point is, a different training/testing set division gives rise to a different classifier-and also to a different estimate of future performance. This is particularly serious if the number of pre-classified examples is small.

Suppose we want to compare two machine-learning algorithms in terms of the quality of the products they induce. The problem of non-representative training sets can be mitigated by the so-called random sub-sampling. 1 The idea is to repeat the random division into the training and testing sets several times, always inducing a classifier from the $i$-th training set, and then measuring the error rate, $E_{i}$, on the $i$-th testing set. The algorithm that delivers classifiers with the lower average value of $E_{i}$ ‘s is deemed better-at least as far as classification performance is concerned.

## 统计代写|机器学习作业代写machine learning代考|Problems with Available Data

The class recognition task, schematically represented by Fig. 1.3, is the most popular task of our discipline. Many concrete engineering problems can be cast in this framework: recognition of visual objects, understanding natural language, medical diagnosis, and identification of hidden patterns in scientific data. Each of these fields may rely on classifiers capable of labeling objects with the right classes based on the features, traits, and attributes characterizing these objects.

Origin of the Training Examples In some applications, the training set is created manually: an expert prepares the examples, tags them with class labels, chooses the attributes, and specifies the value of each attribute in each example. In other

domains, the process is computerized. For instance, a company may want to be able to anticipate an employee’s intention to leave. Their database contains, for each person, the address, gender, marital status, function, salary raises, promotions-as well as the information about whether the person is still with the company or, if not, the day they left. From this, a program can obtain the attribute vectors, labeled as positive if the given person left within a year since the last update of the database record.

Sometimes, the attribute vectors are automatically extracted from a database and labeled by an expert. Alternatively, some examples can be obtained from a database and others added manually. Often, two or more databases are combined. The number of such variations is virtually unlimited.

But whatever the source of the examples, they are likely to suffer from imperfections whose essence and consequences the engineer has to understand.

## 统计代写|机器学习作业代写machine learning代考|Training Sets and Classifiers

AND （填充尺寸=厚）和（填充阴影=深色）

## 广义线性模型代考

statistics-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。