What I really liked about your videos are inclusion of examples along with theory.It is really helpful since we can get a clear picture about what their real use is.
I took the entire Micromaster on EDX, which was really useful (I highly recommend it), but going through your playlists really completes the understanding of all the topics. For example, this was the first time I heard why we care about all the distributions, the actual use of them. Great job!
Hello Sir, You are doing a great job. Your assistance is invaluable, and your concepts are clear, making them easy to learn and remember. Thank you so much for your generous support!😇
I think the mathematics concepts are explained well enough but I don't know why we need to scale our data. I mean the distribution and the conversion is okay but it would have been better if you explained the need for scaling too. anyways @Krish Naik, it's wonderful to see how active you are for your users. if I join you via paid subscription, will I able to ask my doubts and receive answers from you or anyone else? if that is the case please let me know. I'm already very thankful to you!
Great Explanation Sir! I have a doubt, how to check mathematically if data is log normally distributed or normally distributed. because for log normal if I check through visualization (distribution curve), i may end up saying that it is right skewed data with positive outliers.
Good explanation. In the future if u can use a different color it would be great. Dark red on black background is not easy to see. Quick question. How does one decide whether one should apply log or not. You said marketing is log distributed. In reality how does one decide when to use log and when not to use? Thank you
Sure I will make sure about the background color. There is. Concept called as Q-Q plot which helps to find whether the distribution is Log normally distributed
I want to add that the log normal distribution has no negative values, the normal distribution has. And I want to ask what is the expected mean value to the log normal distribution, the arithmetic mean is the expected mean value to the normal distrbution.
calculate the Z score = (value - mean)/ SD gives you the Standardised values. Refere this document : www.mathsisfun.com/data/standard-normal-distribution.html
Very well explained. Just one thing, please use light colour background if you're using dark colour pen or if you're using black background please use light colour pen also please increase your pen's size a bit. It's very difficult to see properly. Content is obviously very good. Please keep posting. Thankyou.
When you mentioned in the example about the Marketing variable following normal distribution and that you assumed you knew this info from domain knowledge, should I understand that this also will be the case if let's say, my sample is not following a normal distribution when I do an exploratory data analysis but I know from domain domain knowledge that my variable is following a normal distribution? Thanks for the great content.
Someone please help me, Here you have converted the input data into different scales and fitting into a model. After training a model, how do you pass the prediction data and get back to initial scale??
Hi krish, I am a bit confused. Suppose one of my variable is not normally distributed, so I apply log() to this to make it normally distributed. Now after applying log(), the variable data points are right skewed. Is this a issue for me or not because I was of the impression that all our features needed to be normally distributed, and not right/left skewed.
Is it compulsory we need to find log normal distribution for market, because if we directly convert market to SD, we will still be able to get data with 0 mean and 1 SD.
Hi Krish, I'm bit confused right now if scaling affect accuracy or not, as per my knowledge it only reduces the training time. Yes, it does have some effect on rmse when we do regression but I'm not sure how accuracy gets affected, please illuminate me on this.
The algorithm (linear or logistic) that we apply understands normally distributed data easily. Basically, we are making things easier for the algorithm and therefore we get better prediction accuracy, because our algorithm is Happy.
@@anubhavsood1510 I know how the algorithm functions and I've checked and verified my professors that there is no impact on the accuracy of the model whether you scale or you don't scale, yes it will reduce the training time so we not making any algorithm happy but we're taking an extra step that after the prediction we need to take the inverse to get the exact value and accuracy I'm talking w.r.t to classification and not regression.
data has to be in normal distribution, only then it makes sense to scale it down using standard scaler or any other method. If you directly apply standard scaler to log normal distributed data, it will have skewness, then you are scaling down data having skewness which is not recommended as told by krish sir.
Hi Krish, Immense pleasure to find one stop solution for statistics learning. Query - Even height of the people could be Normal Distribution as people with height more than 7 feet would be lesser and lesser. Is this true understanding? Thank you Krish!
There is a difference between normal distribution and symmetrical distribution. Every symmetrical distribution is a normal distribution but the vice versa is not TRUE
I told all my Indian engineers: 1) speak slowly, few people understand them 2) speak clearly. But when I watch Indian engineers TH-cam, I have a hard time Chinese engineers do not have this problem of communication. Then, this is my advise to Indian You tubers: Spend some time on English speaking Indian jokes in this country are mostly related to Indian accent
1:30 Log Normal Distribution
4:37 How log normal distribution looks like
9:36 Importance of log distribution
What I really liked about your videos are inclusion of examples along with theory.It is really helpful since we can get a clear picture about what their real use is.
without releating with example. concept going to evaporate in thin air
Thank you Sir for explaining this, Special thanks for explaining "Why we use different ditributions in ML" and its practical implementations.
Thanks a ton ,explanation was crisp and as you had explained with an example ,its easy to connect the dots.Please continue the great work.
thanks again Krish .Once again superb exlanation. Now I understand why do we do normalisation
I took the entire Micromaster on EDX, which was really useful (I highly recommend it), but going through your playlists really completes the understanding of all the topics. For example, this was the first time I heard why we care about all the distributions, the actual use of them. Great job!
Krish, your method of teaching is amazing. Please keep on doing such a great work.
Hello Sir,
You are doing a great job. Your assistance is invaluable, and your concepts are clear, making them easy to learn and remember. Thank you so much for your generous support!😇
Thank you friend :) This explained lognormal in a crisp way for me :)
Learned a lot in 15 mins. Thanks !
Beautifully explained as easy as possible. Thank you so much sir.
What a great explanation. Many many thanks!
Krish, I liked the video before watching it by mistake. Now I've watched the video and I'm glad I liked it.
Amazing explanation Krish. Thanks a ton !!!
I think the mathematics concepts are explained well enough but I don't know why we need to scale our data. I mean the distribution and the conversion is okay but it would have been better if you explained the need for scaling too.
anyways @Krish Naik, it's wonderful to see how active you are for your users. if I join you via paid subscription, will I able to ask my doubts and receive answers from you or anyone else?
if that is the case please let me know.
I'm already very thankful to you!
You did an incredible job keep it up dude!
Thank you krish, this is simply brilliant and outstanding way of explanation. I now remember the days when I learned this topics in my High school.
Awesome Work Krish.. simply explained like a pro :-)
Best explanation Sir, understood complete concept 😊
u explain very nice, liked ur video within 2 minutes of watching,
Great Explanation Sir!
I have a doubt, how to check mathematically if data is log normally distributed or normally distributed.
because for log normal if I check through visualization (distribution curve), i may end up saying that it is right skewed data with positive outliers.
Q-Q plot
You are great sir, keep uploading the videos..
Good explanation. In the future if u can use a different color it would be great. Dark red on black background is not easy to see.
Quick question. How does one decide whether one should apply log or not. You said marketing is log distributed. In reality how does one decide when to use log and when not to use? Thank you
Sure I will make sure about the background color. There is. Concept called as Q-Q plot which helps to find whether the distribution is Log normally distributed
@@krishnaik06 Krish ,same question stricken in my mind .If you have no lecture on Q-Q please make a session.
SIr, there was no discussion about Standard Normal Distribution in the previous video..
Great explanation brother.
Very nice explanation .. thanks
Great Video!!! Really easy to understand
I want to add that the log normal distribution has no negative values, the normal distribution has. And I want to ask what is the expected mean value to the log normal distribution, the arithmetic mean is the expected mean value to the normal distrbution.
Krish,
SND and scale up and scale down you didn't teach in the previous video.
Could you please add a short video for that ?
calculate the Z score = (value - mean)/ SD gives you the Standardised values.
Refere this document : www.mathsisfun.com/data/standard-normal-distribution.html
Awesome explaination
Very clearly explained
Very well explained. Just one thing, please use light colour background if you're using dark colour pen or if you're using black background please use light colour pen also please increase your pen's size a bit. It's very difficult to see properly. Content is obviously very good. Please keep posting. Thankyou.
Do we need to redo the process of scaling before calculating accuracy of the model?
Similar to parents, Teachers are equal to gods. i believed it after watching your channel. Your are very awesome keep rocking.
very easy to understand
Sir why do we need to transform into log and then standard normal distribution. We can right away go with the snd right.
Hi! how do I find the lambda of the product or division of 3 lognormal variables?
Great explanation as always!!
When you mentioned in the example about the Marketing variable following normal distribution and that you assumed you knew this info from domain knowledge, should I understand that this also will be the case if let's say, my sample is not following a normal distribution when I do an exploratory data analysis but I know from domain domain knowledge that my variable is following a normal distribution? Thanks for the great content.
thank you sir.good explanation
Thanks Krish
I think you forgot to discuss SND 🙄
Yeah.I also wonder what difference it with GD
Someone please help me,
Here you have converted the input data into different scales and fitting into a model.
After training a model, how do you pass the prediction data and get back to initial scale??
If it falls log nomal then we can use boxcox method to get std normalization sir
Very helpul.
Thanks
Hi krish,
I am a bit confused. Suppose one of my variable is not normally distributed, so I apply log() to this to make it normally distributed.
Now after applying log(), the variable data points are right skewed. Is this a issue for me or not because I was of the impression that all our features needed to be normally distributed, and not right/left skewed.
Well explained..
great video, one suggestion, please use other color pen (red is difficult to follow)
Super clear. Thank you!!!!
Thank you 😊❤️
Sir how come we find that this data following normal distribution or log normal which disribution this data follow
Is it compulsory we need to find log normal distribution for market, because if we directly convert market to SD, we will still be able to get data with 0 mean and 1 SD.
But sir where have you discussed about Standard normal distribution?? Please help:)
thank you so much!! very helpful
before understanding this, we have to clear the concept related to Standard and Standard ND
What if my target follows a log-normal distribution?
sir by looking at the random variable, how to identify whether it's gaussian,log normal ,binomial or bernoulli distribution?
By Visualization
make one video with example with simple numbers in excel sheet to explain distributions
well explained....
super explanation
Very nice video
Hi Krish,
I'm bit confused right now if scaling affect accuracy or not, as per my knowledge it only reduces the training time. Yes, it does have some effect on rmse when we do regression but I'm not sure how accuracy gets affected, please illuminate me on this.
The algorithm (linear or logistic) that we apply understands normally distributed data easily.
Basically, we are making things easier for the algorithm and therefore we get better prediction accuracy, because our algorithm is Happy.
@@anubhavsood1510 I know how the algorithm functions and I've checked and verified my professors that there is no impact on the accuracy of the model whether you scale or you don't scale, yes it will reduce the training time so we not making any algorithm happy but we're taking an extra step that after the prediction we need to take the inverse to get the exact value and accuracy I'm talking w.r.t to classification and not regression.
krish make some videos of these in python by giving some real time example so that it will be usefull for our interview
Hi Krish why d o we need to apply log. Why can't we directly apply standard scaling?
Having the same doubt. Anyone clarify?
data has to be in normal distribution, only then it makes sense to scale it down using standard scaler or any other method. If you directly apply standard scaler to log normal distributed data, it will have skewness, then you are scaling down data having skewness which is not recommended as told by krish sir.
thank you so much sir
Thank you sir
Your style is like Khan Academy Sal Khan .... repeating the words when writing it.
What is Domain Knowledges?
If i get random data then how i can distribute it can you please guide me
Thanks sir
Hi Krish,
Immense pleasure to find one stop solution for statistics learning.
Query - Even height of the people could be Normal Distribution as people with height more than 7 feet would be lesser and lesser.
Is this true understanding? Thank you Krish!
There is a difference between normal distribution and symmetrical distribution.
Every symmetrical distribution is a normal distribution but the vice versa is not TRUE
SND where??
Great!
if u change the color to white it would be great ...
yes excatly my quation is 10000000000000000000 percent why we learng about distubution ? what we unstnd from distubution ?
except your handwriting everything is awesome
❤💫...
I told all my Indian engineers:
1) speak slowly, few people understand them
2) speak clearly.
But when I watch Indian engineers TH-cam, I have a hard time
Chinese engineers do not have this problem of communication.
Then, this is my advise to Indian You tubers:
Spend some time on English speaking
Indian jokes in this country are mostly related to Indian accent