Today i become a big fan of your lectures... Hi i am following your lectures since last few months and i like them from the beginning, the way how you explain in very simple manner, the technique how to relate all the theory with real world examples, etc... You really doing a fantastic job... You really know how to explain maths in a very common language so it doesn't only fit in my mind but also touches my heart... Thank you for sharing your knowledge with us... I really want to learn more and more with you in near future... You seriously provide a training to the current teachers how to teach the things and how to generate the intrest of learners in any topic... 🥰 Best wishes
@@UnfoldDataScience Thank you actually for sharing your knowledge. I am a data scientist myself, I regularly search TH-cam for quality education. Kudos for the work 😌
of course, it was a great effort to explain PCA in a simple way. I would say at the end of the tutorial you should show the two-way plot explaining the information we are getting from the PCA which was difficult to predict while just looking at the data. Just a suggestion.
You did it better, and few of them are need more clarification especially for bigger datas having 100 or more columns, and how we can rotate the axis by which terms
Jai ho Gurudev ! Sakshat Saraswati ka vaas hi apke kanth me ! Very well explained....one questions. How it gets decided that how much data is explained by PCA1 and how much data has explained by PCA2 and so on ?
Thanks Aman ! Well explained as always. This was my demand few days back and you created this video for all of us once again thanks for this. I have one question if we convert the data to mean centric and taking the covariance matrix what is the intuitions behind this ? Somewhere I read that eigenvector are those vector whose direction does not change when we scaling the matrix so after getting the covariance matrix we are looking that covariance vector whose direction does not change after scaling the data and all those vector are principal component of that data. Please clarify my doubt and correct my understanding.
Thansk Sushant, Think like this. Make data mean centric ( so that covariance matrix is not screwed much even if data columns are on different scale consider milegae of car and it's cost in INR as two different coulmns) Calculate covariance matrix ( just to understand relationship between variables) Find Eigen value and eigen vector( to know on which direction maximum variance is there, may be 1,2,3 any number of directions, as I showed as V1, V2 in matrix example) Once we know in which direction/directions, maximum variance is there, we don't care about covariance matrix anymore, we just take our original data to that direction, we can say project original data to that direction to reduce dimension)
While explaining Eigen value you expanded the matrix like determinant without telling that you are using determinant expansion as matrix can’t be expanded like this this-
Kindly come to basics like on which type of variables PCA is applied. Why not other methods. How to deal with variables having different scales. Everything should start from basics which I found every where missing
Great explanation thanks. Also I have a question; On my dataset 2 features has 0.8 corelation if I use PCA them to decrease one column is it handle 2 features without losing information ? Or should I just drop one column ?
Suppose if I say n=1, n=2, it means we want those many principal components. If you don't pass this argument at all in sklearn, all component are kept which will be equal to no of feature
I have one question, PC1 shows more percentage, which means it should strongly correlate with the original output data. If possible, please clear this doubt.
For creating components, you can create all components, you can keep it default(equal to number of features) however, for choosing how many components for the next step, we see how many "minimum" Components can explain "maximum" Variance together. Let's say PC1 explains 80% variance PC2 explains 15% variance And rest All PC together explain remaining 5% of variance. In this case, we will choose only first two components, PC1 and PC2 for the next step. Just like we choose optimal number of K in K means cluster using elbow method.
Can i use PCA to identify Climate smart Agriculture practices mainly used (adopted) by Household in the study area? pls help how can it is possible. Eg. i have 1.Conservation agriculture (Reduced tillage, Crop residue management-mulching, Crop-rotation/intercropping with cereals and legumes): 2.ISFM (Compost and manure management, Efficient fertilizer application techniques) 3...
Hi, I used XLSTAT and PAST tool to calculate PCs. I need "Contribution of the variables (%)" which I could get in XLSTAT easily but in PAST, I got value of "% variance". Is "% variance" in PAST is same as "Contribution of the variables (%)" in XLSTAT? Please respond. Thanks.
I have one little doubt in python . If interviewer ask tell me about data types in python. Then what exactly we have to told . In our answers how I start . Can I start to data structure or start with saying numeric, logical, ....
You can say simple data type like string, number, Boolean Then come list, array, dict, set Then comes some specific data structure like namedtuple etc. Read about collection module.
you took and mentioned 2 by 2 matrix. but data u took for python is 3 by 2 (three students and two subjects). This cretes confusion. A is not square now. And you first showed plot of original data. Plots after PCA not shown in video. Please show these for better understanding.
Access English, Hindi Course here - www.unfolddatascience.com/store
Don't forget to register on the website, it's free🙂
Code link please
Underrated channel for machine learning god bless you Aman
Dude, you are a really good teacher, awesome methodology!!!
One of the best videos that I have come across for PCA . Thanks @UnfoldDataScience
Weight is vector, mass is scaler. However explained in detail. Great work.
a piece of jargon there
sir thank yoU to clear this concept coz i have been in youtube since 2 hour understanding pca and after watching this video i am clear my doubt
Glad that it was helpful 😊
Thunbs up with 2 hands . Was never able to understand this concepot before.Big Thank you :)
perfectly explained aman thank you!
really good explanation
Awesome lecture.
Better than so called professors
It's my pleasure. Please share with friends
Best video on PCA....keep it up
Thanks Ashwani
Absolutely underrated tutor.
That's a detailed course thanks.
Today i become a big fan of your lectures... Hi i am following your lectures since last few months and i like them from the beginning, the way how you explain in very simple manner, the technique how to relate all the theory with real world examples, etc... You really doing a fantastic job... You really know how to explain maths in a very common language so it doesn't only fit in my mind but also touches my heart... Thank you for sharing your knowledge with us... I really want to learn more and more with you in near future... You seriously provide a training to the current teachers how to teach the things and how to generate the intrest of learners in any topic... 🥰 Best wishes
This is precious
Thanks Aman for such an awesome explanation for a confusing topic like PCA.
Thanks and welcome
The way you explained the vectors mathematically correlated with flight example was wonderful.... 🥰 🎉
Your comment mean a lot to me. Welcome onboard to UFDS
Unbelievable explanation. Wow!!!!
Thanks alot for your positive feedback. Please share with others as well so that everyone gets the knowledge.
Wow . That was just superb.👏👏👏👏
Thank you so much 😀
Thanks aman...
It was really a helpful video.
Thanks Aman for this wonderful explanation
Thank you Anita.
Thanks for the amazing content Aman.
Thanks Amar.
Great explanation.. thank you so much 🎉❤
good job simple and clear understanding
Thanks boss ... Really appreciated .. Good work
You have an outstanding explanation for PCA. All the technical jargon out there is only to confuse people. Cheers.
Thanks for watching, your comments mean a lot.
@@UnfoldDataScience Thank you actually for sharing your knowledge. I am a data scientist myself, I regularly search TH-cam for quality education. Kudos for the work 😌
Thanks for the video, it's too good.
Most welcome
love u sir for this support
Great explanation 👌
Thanks Radha.
Thanks Aman Nicely Explained 🙂👍
Welcome Rohit. pls share with friends who may be interested.
Very nice explanation😊
Great video Aman as usual expected.
Thanks Sambit.
of course, it was a great effort to explain PCA in a simple way. I would say at the end of the tutorial you should show the two-way plot explaining the information we are getting from the PCA which was difficult to predict while just looking at the data. Just a suggestion.
Appreciate your suggestion Rahul. Thanks for watching
very clear and valuable
than you so much,love you bro
Thanks a lot Aman!! well explained 🙂
Very well explained... Thank you very much...
Welcome upendra
@@UnfoldDataScience ❤
I first hit like on your videos and then watch...coz i know you are always awesome 🙂
Thanks Akash 🙂
Thanks Aman👍🙏
Thank you so much for the detailed explanation. Really loved the way you covered each individual basic topic building up to the main topic.
Thanks Manoj.
Great video. Request you to make more videos from basics for the entire data science project lifecycle.
Thank you so much sir ....🎉
Great Videos Aman
Thanks Shaelander
Your explanation is incredible!!!! 👏
Thank you! 😃
great video sir your explanation is amazing🔥
You do a fantastic job explaining complex topics. Definitely subbing
Thanks and welcome
Amazing video, thanks for sharing 🙂
Welcome
You did it better, and few of them are need more clarification especially for bigger datas having 100 or more columns, and how we can rotate the axis by which terms
Nice session 👌
Thanks Varsha.
I was hanging around until I find this video. Thank you sir!
Great video sir
Thank you Jaswanth
Very Nice Explanation. You will never disappoint us 😄
Thanks Adithya.
Weight is not scalar; it's mass pointing towards the direction of gravity. Mass is scalar.
Jai ho Gurudev ! Sakshat Saraswati ka vaas hi apke kanth me ! Very well explained....one questions. How it gets decided that how much data is explained by PCA1 and how much data has explained by PCA2 and so on ?
check EVR (explained variance ratio)
You have explained well but beginners are not able to undersatand the coding phase
It's very understandable
Thanks Again
Very nicely explained 👌. Will be good if a Playlist is created for all ML algo explanations
Sure Kishore.
@@UnfoldDataScience thank you
Please cover data mining, regression, correction, time series
Let me check on these topics, regression and time series playlist are there. You can check in playlist section.
I wnt want to know more about entity embedding for categorical variables as like this
Waiting for videos on LDA , MDS ,t-SNE and PcoA
Yes on the way, thanks for watching.
Thanks a lot.
You're welcome Nehal
Thanks Aman !
Well explained as always. This was my demand few days back and you created this video for all of us once again thanks for this. I have one question if we convert the data to mean centric and taking the covariance matrix what is the intuitions behind this ? Somewhere I read that eigenvector are those vector whose direction does not change when we scaling the matrix so after getting the covariance matrix we are looking that covariance vector whose direction does not change after scaling the data and all those vector are principal component of that data.
Please clarify my doubt and correct my understanding.
Thansk Sushant,
Think like this.
Make data mean centric ( so that covariance matrix is not screwed much even if data columns are on different scale consider milegae of car and it's cost in INR as two different coulmns)
Calculate covariance matrix ( just to understand relationship between variables)
Find Eigen value and eigen vector( to know on which direction maximum variance is there, may be 1,2,3 any number of directions, as I showed as V1, V2 in matrix example)
Once we know in which direction/directions, maximum variance is there, we don't care about covariance matrix anymore, we just take our original data to that direction, we can say project original data to that direction to reduce dimension)
Also those vectors are not principal components, once u project your original data to that vector direction then u get principal components
Thanks ! Got it.
Sir please make video on exploratory data analysis
While explaining Eigen value you expanded the matrix like determinant without telling that you are using determinant expansion as matrix can’t be expanded like this this-
Kindly come to basics like on which type of variables PCA is applied. Why not other methods. How to deal with variables having different scales. Everything should start from basics which I found every where missing
Can you logic behind how to calculate Variance Explanantion by each PCA component? Keep up the good work. Thanks
Hi aman,
Can you please explain about quantization aware training, why it is used compared to floating point model
Thanks Sharan, I will try to bring video on it.
Can you please make a video on OLPP
Pls use presentation mode in jupyter so I can view code fonts large in mobile, thnks
Sure thanks
Can u help with regularised k means clustering
Great explanation thanks. Also I have a question; On my dataset 2 features has 0.8 corelation if I use PCA them to decrease one column is it handle 2 features without losing information ? Or should I just drop one column ?
thank you
Is it possible to have an example of pictures to classify them into two categories?
You're welcome
sure
@@UnfoldDataScience If the dimensions are reduced in pca and classification in knn is better , please
hi
Thank you sir. It is great video. Just one thing need to know, incase of PCA also, we need to do data cleaning or directly we can proceed for PCA??
Data cleaning will help PCA create more meaningful contents.
What happen if we don don't pass n_components argument
Suppose if I say n=1, n=2, it means we want those many principal components. If you don't pass this argument at all in sklearn, all component are kept which will be equal to no of feature
@@UnfoldDataScience How do you decide what should be the optimal number for "n" ?
I have one question, PC1 shows more percentage, which means it should strongly correlate with the original output data. If possible, please clear this doubt.
Hello Aman, Nice Explanation. but one question is it necessary all data set go through PCA or when we will use PCA
How do we come up with the number for "n_components"?
For creating components, you can create all components, you can keep it default(equal to number of features) however, for choosing how many components for the next step, we see how many "minimum" Components can explain "maximum" Variance together.
Let's say PC1 explains 80% variance
PC2 explains 15% variance
And rest All PC together explain remaining 5% of variance.
In this case, we will choose only first two components, PC1 and PC2 for the next step.
Just like we choose optimal number of K in K means cluster using elbow method.
@@UnfoldDataScience am just reading your comment while watching your Regularisation video. Thank you sooo much. ♥️
Can i use PCA to identify Climate smart Agriculture practices mainly used (adopted) by Household in the study area? pls help how can it is possible. Eg. i have 1.Conservation agriculture (Reduced tillage, Crop residue management-mulching, Crop-rotation/intercropping with cereals and legumes): 2.ISFM (Compost and manure management, Efficient fertilizer application techniques) 3...
Sir pls make videos in hindi also
Will check the plan, thanks for watching.
Thank you, but why do we do mean centered
What the results tell or what it denote
how are we calculating PC1 after projecting our data to new axis
Sir, how do i label or annotate the data point after clustering. I have used covid 19 data set for pca analysis
Good question, take reference from original data rowwise.
How to find the value of pc1 using python code
Hi, I used XLSTAT and PAST tool to calculate PCs. I need "Contribution of the variables (%)" which I could get in XLSTAT easily but in PAST, I got value of "% variance". Is "% variance" in PAST is same as "Contribution of the variables (%)" in XLSTAT? Please respond. Thanks.
I have one little doubt in python . If interviewer ask tell me about data types in python. Then what exactly we have to told . In our answers how I start . Can I start to data structure or start with saying numeric, logical, ....
You can say simple data type like string, number, Boolean
Then come list, array, dict, set
Then comes some specific data structure like namedtuple etc. Read about collection module.
@@UnfoldDataScience Thank you sir
Can we do PCA on the combined results of samples from two separate distributions?
you took and mentioned 2 by 2 matrix. but data u took for python is 3 by 2 (three students and two subjects). This cretes confusion. A is not square now. And you first showed plot of original data. Plots after PCA not shown in video. Please show these for better understanding.
Can you provide the source code..
Link in description
14:37
Well Explained.......Thank You!
Thanks
Welcome