Logistic Regression – Geometric Intuition
Logistic Regression – Geometric Intuition
Everybody who has taken a machine learning course probably knows the geometric intuition behind a support vector machine (SVM, great book): A SVM is a large margin classifier. In other words, it maximizes the geometric distance between the decision boundary and the classes of samples. Often you’ll find plots similar to this one:
But what about logistic regression? What is the geometric intuition behind it and how does it compare to linear SVMs? Let’s find out.
Geometric intuition behind logistic regression
First, a quick reminder about the definition of the logistic function, given features:
With that out of the way, let’s dive into the geometric aspects of logistic regression, starting with a toy example of only one feature:
The blue samples of our feature data belong to class , while the red dots are supposed to get assigned to . As we can see in the figure above, our data is fully separable and the logistic regression is able to demonstrate this property. From a geometric perspective, the algorithm essentially adds a new dimension for our dependent variable and fits a logistic function to the arising two-dimensional data in such a way that it separates the samples as good as possible within the range of . Once we have such a fitted curve, the usual method of prediction is to assign to everything and vice versa.
When thinking about logistic regression, I usually have its tight connection to linear regression somewhere in the back of my head (no surprise, both have “regression” as part of their names). Therefore, I was wondering what the line would look like and added it to the plot. As we can easily see, differs significantly from the potential result of a linear regression for the given input data. While that was to be expected, I think it’s still nice to confirm it visually.
Note two details of the plot: For one, the logistic curve is not as steep as you might expect at first. This is due to the use of regularization, which logistic regression basically can’t live without. Second, for and not, as my intuition initially suggested, for . Not sure if this is particularly helpful, but I find it interesting nonetheless.
So far, so good. But what happens when we introduce an additional dimension? Let’s have a look at such a two feature scenario (feel free to pinch and zoom the plot):
What we see is the surface of the fitted logistic function for a classification task with two features. This time, instead of transforming a line to a curve, the equivalent happened for a plane. To make it easier to understand how to carry out predictions given such a model, the following plot visualizes the data points projected onto the surface of the logistic regression function:
Now we can again make decisions for our class based on a threshold , which usually is :
So much about the geometric intuition behind logistic regression. Let’s see how that compares to linear SVMs.
Comparison of logistic regression to linear SVM
First of all, notice that with logistic regression we can use the value of a sample as a probability estimate for . The responsive 3D-plots above created with the help of plotly’s fantastic library illustrate this property of logistic regression very well. I have yet to come across a real-world classification task where such probability estimates are not at least somewhat useful. Alas, SVMs (in their original version) aren’t able to provide them. Due to the way the model works, all it can tell you is whether a sample belongs to class or not.
I think at this point the most effective way of comparing logistic regression to linear SVM geometrically, is to add the decision boundary of logistic regression to the initial figure of the post. To make this happen, we need to project the decision boundary of the three-dimensional plot above onto the two-dimensional feature space. Doing so, we end up with the following figure:
As we can see, the logistic regression does not follow the same large margin characteristic as a linear SVM. I’ll leave it to you to decide which is the better fit for the given data.
If you’d like to know further differences between logistic regression and linear SVMs aside from the geometrically motivated ones mentioned here, have a look at the excellent answers on quora.
The code for generating the plots of this post can be found on github.
Logistic Regression – Geometric Intuition的更多相关文章
- 课程一(Neural Networks and Deep Learning),第二周(Basics of Neural Network programming)—— 4、Logistic Regression with a Neural Network mindset
Logistic Regression with a Neural Network mindset Welcome to the first (required) programming exerci ...
- 机器学习 (三) 逻辑回归 Logistic Regression
文章内容均来自斯坦福大学的Andrew Ng教授讲解的Machine Learning课程,本文是针对该课程的个人学习笔记,如有疏漏,请以原课程所讲述内容为准.感谢博主Rachel Zhang 的个人 ...
- 逻辑回归 Logistic Regression
逻辑回归(Logistic Regression)是广义线性回归的一种.逻辑回归是用来做分类任务的常用算法.分类任务的目标是找一个函数,把观测值匹配到相关的类和标签上.比如一个人有没有病,又因为噪声的 ...
- logistic regression与SVM
Logistic模型和SVM都是用于二分类,现在大概说一下两者的区别 ① 寻找最优超平面的方法不同 形象点说,Logistic模型找的那个超平面,是尽量让所有点都远离它,而SVM寻找的那个超平面,是只 ...
- Logistic Regression - Formula Deduction
Sigmoid Function \[ \sigma(z)=\frac{1}{1+e^{(-z)}} \] feature: axial symmetry: \[ \sigma(z)+ \sigma( ...
- SparkMLlib之 logistic regression源码分析
最近在研究机器学习,使用的工具是spark,本文是针对spar最新的源码Spark1.6.0的MLlib中的logistic regression, linear regression进行源码分析,其 ...
- [OpenCV] Samples 06: [ML] logistic regression
logistic regression,这个算法只能解决简单的线性二分类,在众多的机器学习分类算法中并不出众,但它能被改进为多分类,并换了另外一个名字softmax, 这可是深度学习中响当当的分类算法 ...
- Stanford机器学习笔记-2.Logistic Regression
Content: 2 Logistic Regression. 2.1 Classification. 2.2 Hypothesis representation. 2.2.1 Interpretin ...
- Logistic Regression vs Decision Trees vs SVM: Part II
This is the 2nd part of the series. Read the first part here: Logistic Regression Vs Decision Trees ...
随机推荐
- 如何在内网安装compass
神器compass是肿么用这里不做介绍,因为我也不清楚,可参考官网:http://compass-style.org.这里主要介绍如何在内网安装compass. 首先介绍一般是如何安装compass的 ...
- Hadoop框架
1.Hadoop的整体框架 Hadoop由HDFS.MapReduce.HBase.Hive和ZooKeeper等成员组成,其中最基础最重要元素为底层用于存储集群中所有存储节点文件的文件系统HDFS( ...
- Jenkins 构建运行java程序
我们将在Jenkins建立执行一个简单的 HelloWorld 应用程序,构建和运行Java程序.打开网址:http://localhost:8080/jenkins 第1步- 转到Jenkins 仪 ...
- linux第十八章学习笔记
第十八章 调试 内核级开发的调试工作远比用户级开发艰难,它带来的风险比用户级别更高. 一.准备开始 1. 准备工作需要: 一个bug 一个藏匿bug的内核版本 相关内核代码的知识和运气 2. 在用户级 ...
- XML中<beans>属性
<beans xmlns="http://www.springframework.org/schema/beans" xmlns:xsi="http://www.w ...
- Alpha冲刺——day3
Alpha冲刺--day3 作业链接 Alpha冲刺随笔集 github地址 团队成员 031602636 许舒玲(队长) 031602237 吴杰婷 031602220 雷博浩 031602634 ...
- linux下MySQL使用方法
一.登录MySQL 登录MySQL的命令是mysql, mysql 的使用语法如下: mysql [-u username] [-h host] [-p[password]] [dbname] us ...
- Berland and the Shortest Paths CodeForces - 1005F(最短路树)
最短路树就是用bfs走一遍就可以了 d[v] = d[u] + 1 表示v是u的前驱边 然后遍历每个结点 存下它的前驱边 再用dfs遍历每个结点 依次取每个结点的某个前驱边即可 #include &l ...
- 【Revit API】创建工作集并将element加入工作集中
话不多说,直接上代码! public class WorkSetHelper { public void AddElementsToWorkSet(Document doc, List<Elem ...
- 解决 No Entity Framework provider found for the ADO.NET provider
方法很简单,添加下面的dll即可 EntityFramework.SqlServer.dll 疯吻IT