Very well explained sir. Just to add the entire code at 11:15 dataframe = pd.DataFrame(data= new_coordinates, columns= ("1st_Principal", "2nd_Principal", "labels"))
The reason why the 2 plots (your plot and the one from scikit-learn library) are not the same is that you are not reversing the vectors array along the columns (vectors = vectors[:, ::-1]), since eigh() is returning the vectors on ascending order and you need them on descending order.
As said in the appliedAI website under this video description: correction: The formula for covariance matrix is 1/n * (X^T*X) (we missed out in 1/n in the video) there is a typo in the ipython notebook, as eigenvalues generated are in ascending order, when we multiply vector*sample_data^T vector[0]*X[i] will be second principle component vector[1]*x[i] will be first principle component
That's just a Google search away. Just search for " MNIST dataset download" and you will get multiple sources for this data. This dataset is also inbuilt in most libraries like Scikit Learn.
Very well explained sir.
Just to add the entire code at 11:15
dataframe = pd.DataFrame(data= new_coordinates, columns= ("1st_Principal", "2nd_Principal", "labels"))
can you please provide the code
if you have?
The reason why the 2 plots (your plot and the one from scikit-learn library) are not the same is that you are not reversing the vectors array along the columns (vectors = vectors[:, ::-1]), since eigh() is returning the vectors on ascending order and you need them on descending order.
One of the best explaination on PCA so far on RUclips thanks a lot ❤️
why1/n is not multiplied with covar_matrix? Is it done in fit_transform(data) function?
As said in the appliedAI website under this video description:
correction: The formula for covariance matrix is 1/n * (X^T*X) (we missed out in 1/n in the video) there is a typo in the ipython notebook, as eigenvalues generated are in ascending order, when we multiply vector*sample_data^T vector[0]*X[i] will be second principle component vector[1]*x[i] will be first principle component
Why you did the transpose you could have performed matmul(sample_data,vectors) ?
can you please upload the code by any link
Should we have to add labels according to row manually?
where we are setting the number of principal components in the first expaination ?
in the data frame , the column part should include label as well. you did not show
Straightforward explanation. Thanks for the great content!!
Extremely useful. Thank you!
if you have the code please can you share it?
can we take your code for investigate
sir from where i get the dataset used in this video
please provide link of the dataset
That's just a Google search away. Just search for " MNIST dataset download" and you will get multiple sources for this data. This dataset is also inbuilt in most libraries like Scikit Learn.
where is the github link ?