Mean Square Error

LossFunction , Error , SquareError, LocalMinima, GlobalMinima , Advantage_of_MSE

[@jackripper6066]

This was fucking amazing. I was stuck on this concept for hours and you explained it so simply. Thanks.

[@learndatasciencewithpranja7060]

Thank You Jack for liking the video... I hope you will enjoy other videos also ...

[@anindian4601]

same with me bro, u said correct

[@rupalidatir3666]

Thank You sir for uploading the videos, clearly understand the concept of MSE and had understand y we are using error square and not raise to power 3,4..

[@learndatasciencewithpranja7060]

I have added a second video on the mean square error below is the link kzbin.info/www/bejne/gn25c3Wvm8mnh5o this video will give you understanding why we use square in MSE

[@danishpatel8249]

Very well explained sir, now I am confident enough to face this question in interview 😀😀.. Again thanks for this video.... And specially the last comment in which you explained MAE OR MSE.

[@Hsn-Ab]

Now I got the concept behind the use case of mean square error..earlier I thought we could take the Modulus of all error digits if its only the matter of making a negative number to its respective positive instead of mean square error function ....thank you

[@learndatasciencewithpranja7060]

Thank You Hasan for liking the video

[@prigyashukla1450]

very clear and concise explanation ! thanks a lot !!

[@learndatasciencewithpranja7060]

Thanks Pragya

[@anindian4601]

U R THE BEST TEACHER , PLEASE DO KEEP POSTING

[@learndatasciencewithpranja7060]

Thanks for liking and giving such a encouraging feedback

[@ajaykushwaha4233]

Very well u have explained. Thank you for such a nice video.

[@learndatasciencewithpranja7060]

Thanks Ajay for liking the video...

[@omkarred]

superb explanation !!! Appreciate your video ...

[@learndatasciencewithpranja7060]

Thanks Omkar for liking the video I hope you will like other videos also

[@teresaconners7253]

Very helpful!

[@learndatasciencewithpranja7060]

Thanks Teresa for liking the video...I hope you will enjoy other videos also on the channel

[@patrickbateman7665]

What a Beautiful video. Thank you Sir

[@learndatasciencewithpranja7060]

Thank You Dileep for liking the video. I hope you will enjoy watching other videos also .

[@patrickbateman7665]

@@learndatasciencewithpranja7060 I will definitely look out. Thank you once again for short and brief video 😊🙏

[@ahmedgoharkhan]

thank you

[@alibaba888]

10/10 THANK YOU!!

[@learndatasciencewithpranja7060]

Thank You Kaguya for such encouraging remarks and for liking the video.

[@manasathamatani8614]

Sir, thank you very very much.

[@learndatasciencewithpranja7060]

Thanks Manasa

[@Dr.AhmedQusaySabri]

Thanks a lot

[@learndatasciencewithpranja7060]

Thanks Dr Ahmed for liking the Knowledge Video

[@shahbazdeshmukh1484]

Thank you Sir

[@learndatasciencewithpranja7060]

Thanks Shahbaz

[@GurpreetSingh-es9ni]

Hi Sir, it was really informative video, however i didn't understand the last part of this video. starting from 7.31, can you please make a video on that part.

[@learndatasciencewithpranja7060]

Hi Gurupreet , Thanks for liking the video. i will make a follow up video on the MSE to explain it

[@learndatasciencewithpranja7060]

i have created 1 more video kzbin.info/www/bejne/gn25c3Wvm8mnh5o it will answer your query you have asked

[@user-bz7fj1fk2m]

Then why again RMSE? RMSE =Sq.root of MSE. So why? Isn't MSE enough? Thank U

[@sumeersaifi6354]

x^4 will also have one minima i did not understand what you want to say

[@learndatasciencewithpranja7060]

whether you take square, power 4 or 6 or 8 or 10 or any even power there shape would be same ..but while training the model you take 'Gradient' that is your slope which i hope you understand is dy/dx. when you take slope of square it would be linear if you take slope of power 4 it will be cube and cube(odd power) will have local and global minima. Similarly if original function is power 6 then when you take slope(dy/ dx) it would be power 5( odd power) and again you will have problem of local and global minima I hope you understand now....

[@sreenathgupta6767]

Instead of square, why can't we use absolute value which doesn't nullify the error.

[@learndatasciencewithpranja7060]

Hi Sreenath , MAE ( mean absolute error) can be used and is used to evaluate the performance of the model( i.e how good or bad the model is ) and similarly MSE can also be used to evaluate the performance of the model...and similarly RMSE, MAPE can also be used for model evaluation so now what's the big difference between MAE and MSE difference is MSE is a square ( degree 2 function) and MAE is a degree 1 function 2 huge advantages of MSE are 1> MSE will punish your model for large errors ( this part is already explained clearly in the video ) 2 > MSE can be used as a cost function . Cost function is used to train the model. remember one thing no other matrix ( MAE , RMSE , MAPE ) can be used as the cost function because they are not square and they are single degree function and on single degree function when you find the slope its a constant for example lets say y = 10 + 4x is a function and now if you take the derivative of it it would be 4 which is a constant so now you cant not use the concept of gradient descent on the constant but on another hand if i have a square function like y = 6x*x if you take a derivate of it , it would be y = 12x, so now now i can find the slope and concept of gradient descent can be used to train the model....this can be used to train the model because is get the direction in which the errors will reduce i get the relation ship between y and x.. but if its a degree one equation then as soon as you take 1st derivate it becomes a constant. And the relation ship between X and Y can not be established soon i would be putting a video on gradient descent .. I hope this helps

[@sreenathgupta6767]

@@learndatasciencewithpranja7060 thanks a lot for detailed explanation.

[@gregkumar7478]

Why do we square the difference instead of using the abolute value?

[@learndatasciencewithpranja7060]

Hi Greg , I guess i have clearly explained this point why we take square, example taken in the video Model A and Model B( when the actual value is 13 , and predicted value is 14 in model A and 9 in model B) then when we take the square large error are magnified. Total error in 2 models may be same , but square error shoots up for the large error. So MSE is used to Punish the model for the large errors please revisit the video once more , i guess your query would be cleared Thanks

[@user-bz7fj1fk2m]

​@@learndatasciencewithpranja7060 10QU!!! May God reward u with health and more blessings!!!!!