- Hands-On Mathematics for Deep Learning
- Jay Dawani
- 564字
- 2021-06-18 18:55:25
Continuous random variables
So far, we've looked at discrete outcomes in the sample space where we could find the probability of a certain outcome. But now, in the continuous space, we will find the probability of our outcome being in a particular interval or range.
Now, to find the distribution of X, we need to define a function, f, so that the probability of X must lie in the interval
Formally, a random variable, , is continuous if, in a function,
so that we have the following:

We call the function, f, the probability density function (PDF) and it must satisfy the following:
There is another distribution function that is important for us to know, known as the cumulative distribution function. If we have a random variable, X, that could be continuous or discrete, then, , where F(x) is increasing so that x→∞ and F(x)→1.
When dealing with continuous random variables such as the following, we know that F is both continuous and differentiable:

So, when F is differentiable, then F'(x) = f(x).
An important fact to note is that .
This leads us to the concept of uniform distribution, which, in general, has the following PDF:

So, we have the following:

This is the case for .
We write if X follows a uniform distribution on the [a, b] interval.
Now, let's suppose our random variable is an exponential random variable and has the added λ parameter. Then, its PDF is and
for all
.
We write this as , so we have the following:

It is also very important to note that the exponential random variable, such as the geometric random variable, is memory-less; that is, the past gives us no information about the future.
Just as in the discrete case, we can define the expectation and variance in the case of continuous random variables.
The expectation for a continuous random variable is defined as follows:

But, say . Then, we have the following:

In the case of continuous random variables, the variance is defined as follows:

This gives us the following:

Now, for example, let's take . We can find the expected value of X as follows:

Its variance can be found as follows:

Now, we have a good handle of expectation and variance in continuous distribution. Let's get acquainted with two additional terms that apply to PDFs—mode and median.
The mode in a PDF is the value that appears the most; however, it is also possible for the mode to appear more than once. For example, in a uniform distribution, all the x values can be considered as the mode.
Say we have a PDF, f(x). Then, we denote the mode as , so
for all cases of x.
We define the median as follows:

However, in a discrete case, the median is as follows:

Many times, in probability, we take the sample mean instead of the mean. Suppose we have a distribution that contains all the values that X can take. From it, we randomly sample n values and average it so that we have the following:

- 計算機組成原理與接口技術:基于MIPS架構實驗教程(第2版)
- 數據庫原理及應用教程(第4版)(微課版)
- Unity 5.x Game AI Programming Cookbook
- SQL Server 2012數據庫技術與應用(微課版)
- Enterprise Integration with WSO2 ESB
- 智能數據分析:入門、實戰與平臺構建
- 數據庫原理與應用
- MySQL技術內幕:SQL編程
- 新手學會計(2013-2014實戰升級版)
- 企業大數據處理:Spark、Druid、Flume與Kafka應用實踐
- Oracle 11g數據庫管理員指南
- Hadoop大數據技術開發實戰
- Hive性能調優實戰
- 實用數據結構基礎(第四版)
- 反饋:化解不確定性的數字認知論