Maximum likelihood estimation of Linear Regression | PSN Academy

Finds the best values of parameters for which the model has the best fit i.e. optimized. This finding procedure is based on Probability theory.
You may argue that we can determine the 𝜇 and 𝜎 directly from the dataset. What’s the point of trial and error?
Well that’s true in this case. But there are situations where a direct method of determining the parameter values is not available. In those cases, we have to fit in the values of the parameters to the system using some efficient method. Those predicted values must be as close as possible to the true values i.e., the error between our predicted values and the true values is minimum.
The distribution of this error (each error for each example) is assumed to be normally, independently and identically distributed (i.i.d.).
----------------------------------------
Chapters:
00:09 What is Maximum Likelhoood Estimation?
00:37 Is the method based on Probability framework?
05:49 What is an example of Normal Distribution?
03:07 What are the parameters of Normal Distribution?
06:19 What is Linear Regression?
08:17 What is Bias in Linear Regression?
09:20 What is Rotation operation in Linear Regression?
09:31 What is Shift operation in Linear Regression?
10:34 What is Error in Linear Regression?
12:49 What is the Probability Density Function of one data point?
15:43 What is the Probability Density Function of all points?
17:16 What is Likelihood function?
18:49 What is Log-Likelihood function?
22:37 How to determine w0 in Linear regression?
27:40 How to determine w1 in Linear regression?
35:40 How covariance is equal to sum of xy minus sum of xbar ybar?
38:16 How variance is equal to sum of x squared minus sum of xbar squared?
40:34 How to determine sigma squared in Linear regression?
#linearregression #mle #maximumlikelihoodestimation #variance #covariance #bias #error #normaldistribution #loglikelihood #psnacademy