## 计算机代写|机器学习代写machine learning代考|COMP30027

2022年12月23日

couryes-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代写方面经验极为丰富，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
couryes™为您提供可以保分的包课服务

## 计算机代写|机器学习代写machine learning代考|Nonlinear Regression

So far, we have limited our discussion to models of the form $y=X \theta$, mostly because these offered us a convenient (closed form) solution to finding lines of best fit in terms of $\theta$.

However, this type of model has several limitations that we might wish to overcome, such as:
14 The derivative of Equation (2.54) is more obvious after expanding $x_i \cdot \theta=\sum_{k=1}^K x_{i k} \theta_k$.

• We cannot incorporate simple constraints on our parameters, such as that a certain parameter should be positive, or that one parameter is larger than another (which might be based on domain knowledge of a certain problem).
• Although we can manually engineer nonlinear transforms of our features (as we did in sec. 2.3.1), we cannot have the model learn these nonlinear relationships automatically.
• The model cannot learn complex interactions among features, for example, that length is correlated with ratings, but only if the user is female. ${ }^{15}$

These goals can potentially be realized if we are allowed to transform model parameters: for instance, we could ensure that a particular parameter was always positive by fitting
$$\theta_k=\log \left(1+e^{\theta_k^{\prime}}\right)$$
(this is known as a ‘softplus’ function; note that this function smoothly maps $\theta_k^{\prime} \in \mathbb{R}$ to $\theta_k \in(0, \infty)$ ); or if we wanted one feature to be larger than another (e.g., $\theta_k>\theta_j$ ) we could simply add the positive quantity above to another feature:
$$\theta_k=\theta_j+\log \left(1+e^{\theta_k^{\prime}}\right) .$$
Roughly speaking, fitting these types of nonlinear models (and especially models that deal with complex combinations of parameters) is the basic goal of deep learning. We will see various examples of nonlinear models in later chapters, including models based on deep learning (e.g., secs. $7.6$ and 9.4). In Chapter 3 (sec. 3.4.4) we present the basic approach used to fit these types of models using high-level optimization libraries.

## 计算机代写|机器学习代写machine learning代考|Case Study: Image Popularity on Reddit

Lakkaraju et al. (2013) used regression algorithms to estimate the success of content (e.g., number of upvotes) on reddit. Other than building an accurate predictor, their main goal is to understand and disentangle which features are most influential in determining content popularity.

Presumably, one of the biggest predictors of success is the quality of the content itself. Predicting whether a submission is of high quality (e.g., whether an image is funny or aesthetically attractive) is presumably incredibly challenging. To control for this high-variance factor of content quality, Lakkaraju et al. (2013) study resubmissions, that is, content (images) that has been submitted multiple times. This way, if one submission is more successful than another (of the same image), the difference in success cannot be attributed to the content itself, and must arise due to other factors such as the title of the submission or the community it was submitted to.

Having controlled for the effect of the content itself, the goal is then to distinguish between features that capture the specific dynamics of reddit itself, versus those that arise due to the choice of title (i.e., how the content is ‘marketed’). Various features are extracted that model reddit’s community dynamics, such as the following:

• One of the largest predictors of successful content is simply whether it has been submitted before (as we saw in Figure $2.13$, which is based on the same dataset); this is eaptured via an exponentially decaying function.
• However, the above effect might be mitigated if enough time has passed between resubmissions (by when the original submission is forgotten, or the community has enough new users); this is captured using a feature based on the inverse of the time delta between submissions.
• Resubmissions might still be successful if they are resubmitted to largely non-overlapping communities (subreddits).
• Submission success may correlate with the time of day. For example, submissions may be most successful during the highest-traffic times of day, or alternately they may be more successful if submitted when there is less competition.

Whereas community effects are somewhat reddit-specific, measuring the effect of a particular choice of title can potentially be of broader interest. Understanding the characteristics of successful titles can have implications when marketing content (such as an advertising campaign) to a new market.

# 机器学习代考

## 计算机代写|机器学习代写machine learning代考|Nonlinear Regression

14 方程 (2.54) 的导数在展开后更加明显 $x_i \cdot \theta=\sum_{k=1}^K x_{i k} \theta_k$.

• 我们不能对我们的参数进行简单的约束，例如某个参数应该为 正，或者一个参数大于另一个（这可能基于某个问题的领域知 识)。
• 尽管我们可以手动设计特征的非线性变换 (如我们在第 $2.3 .1$ 节中所做的那样)，但我们不能让模型自动学习这些非线性关 系。
• 该模型无法学习特征之间的复杂交互，例如，长度与评级相 关，但前提是用户是女性。 15
如果允许我们转换模型参数，这些目标就有可能实现: 例如，我们可 以通过拟合确保特定参数始終为正
$$\theta_k=\log \left(1+e^{\theta_k}\right)$$
(这被称为”softplus”功能; 请注意，此功能平滑映射 $\theta_k^{\prime} \in \mathbb{R}$ 到 $\theta_k \in(0, \infty)$ ); 或者如果我们㹷望一个特征比另一个大 (例如， $\theta_k>\theta_j$ ) 我们可以简单地将上面的正数量添加到另一个特征中:
$$\theta_k=\theta_j+\log \left(1+e^{\theta_k^{\prime}}\right) .$$
粗略地说，拟合这些类型的非线性模型（尤其是处理复杂参数组合的 模型) 是深度学习的基本目标。我们将在后面的章节中看到非线性模 型的各种示例，包括基于深度学习的模型 (例如， secs.7.6和 9.4)。在 第 3 章 (第 $3.4 .4$ 节) 中，我们介绍了使用高级优化库来拟合这些类型 模型的基本方法。

## 计算机代写|机器学习代写machine learning代考|Case Study: Image Popularity on Reddit

Lakkaraju 等人。(2013) 使用回归算法来估计 reddit 上内容的成功程度（例如，赞成票的数量）。除了构建准确的预测器之外，他们的主要目标是了解和理清哪些特征对确定内容流行度影响最大。

• 成功内容的最大预测因素之一就是它之前是否已提交（如图所示）2.13，基于相同的数据集）；这是通过指数衰减函数获取的。
• 但是，如果两次重新提交之间间隔足够长的时间（当原始提交被遗忘，或者社区有足够多的新用户时），上述影响可能会减轻；这是使用基于提交之间的时间增量倒数的功能捕获的。
• 如果重新提交给基本上不重叠的社区（subreddits），重新提交可能仍然会成功。
• 提交成功可能与一天中的时间相关。例如，提交可能在一天中流量最高的时间最成功，或者如果在竞争较少时提交，则提交可能更成功。

## 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。