Regularization in machine learning | L1 and L2 Regularization | Lasso and Ridge Regression

Поділитися
Вставка
  • Опубліковано 27 січ 2025

КОМЕНТАРІ • 87

  • @mrutyunjayaraghuwansi4368
    @mrutyunjayaraghuwansi4368 3 роки тому +13

    After my data science classes I used to watch the concepts through your videos and it helped me a lot in understanding... 😃😃

  • @brahmadanna
    @brahmadanna 3 роки тому +9

    Every one can understand ur explanation.....neat and clear 👍

  • @shanmukhchandrayama8508
    @shanmukhchandrayama8508 3 роки тому +4

    Beautiful explanation sir, I regret of not watching this video before my interview but anyhow I am glad I got to know it now.

  • @HimanshuKumar-oi8qh
    @HimanshuKumar-oi8qh 2 роки тому +1

    Thanks ! all doubts cleared ..!
    The word sweet spot can actually impress the interviewer I guess :)

  • @kalyanbikramadhikari6049
    @kalyanbikramadhikari6049 Рік тому

    loved the way u teach and your voice is amazaing. I wish for the growth of this channel

  • @dehumanizer668
    @dehumanizer668 3 роки тому +2

    Excellent teacher. Thank you sir for such a wonderful explanation. :)

  • @sudhanshusoni1524
    @sudhanshusoni1524 3 роки тому +2

    Amazing, your teaching skills are really awesome sir! Thanks for this great work

  • @kirandeepmarala5541
    @kirandeepmarala5541 3 роки тому +1

    Really Learned a lot Sir..your teaching skills are amazing..Super..

  • @GauravExplorations
    @GauravExplorations 3 роки тому

    one of the good explanations i have seen for this topic, good work

  • @sane7263
    @sane7263 Рік тому

    1:09 that laugh 🤣🤣
    I understand the struggle.

  • @christysujatha
    @christysujatha 3 роки тому +1

    Very clear explanation 👍👍👍

  • @PramodKumar-su8xv
    @PramodKumar-su8xv 9 місяців тому

    good explanation with keeping the audience understanding in me

  • @ajitkulkarni1702
    @ajitkulkarni1702 10 місяців тому

    Amazing explaination Sir !!!

  • @shine_through_darkness
    @shine_through_darkness 10 місяців тому +1

    feels like getting a lecture from one of my friends at last night before the exam

  • @inderjeetsingh5959
    @inderjeetsingh5959 3 роки тому

    Thank you sir .
    You just made tuff topics so easy.🙏

  • @Engineer_Boy_01
    @Engineer_Boy_01 7 місяців тому +1

    Thanks vaiya😊

  • @miteshkumarsingh
    @miteshkumarsingh Рік тому

    Very Well explained.

  • @souravbiswas6892
    @souravbiswas6892 3 роки тому +1

    Excellent explanation 👍🏻

  • @yogeshbhatt3806
    @yogeshbhatt3806 Рік тому

    Awsome description

  • @king1_one
    @king1_one 6 місяців тому

    great explanation sir

  • @adarsh5856
    @adarsh5856 2 роки тому

    Amazing Teaching Sir.. Thank You....

  • @navoditmehta8833
    @navoditmehta8833 3 роки тому +1

    Very nicely explained 💯💯
    Please explain the maths behind feature selection using lasso and not ridge.

  • @adwaitkotewar4682
    @adwaitkotewar4682 Рік тому +2

    Sweet Spot ❌ Technical word - Balanced Fit

  • @subhajitdutta1443
    @subhajitdutta1443 2 роки тому

    Great... Helped a lot..

  • @mustafizurrahman5699
    @mustafizurrahman5699 Рік тому

    Awesome sir

  • @spivak4342
    @spivak4342 9 місяців тому

    THANK YOU MR. AMAN SIR

  • @pipicovers
    @pipicovers 2 роки тому

    i wish this channel reached 100K very soon

    • @UnfoldDataScience
      @UnfoldDataScience  2 роки тому

      Thank you so much. If you guys keep liking and sharing, anything is possible. Your feedback is highly appreciated!

  • @keemsterbaas7156
    @keemsterbaas7156 3 роки тому

    Excellent explanation. Subscribed!

  • @sandipansarkar9211
    @sandipansarkar9211 3 роки тому

    fnished watching

  • @homosapien7754
    @homosapien7754 3 роки тому +3

    Wouldn't cubing the slope(instead of squaring) in the ridge regression penalty decrease the loss function even more? If yes, why don't we do that?

    • @SoumyaDasgupta
      @SoumyaDasgupta 3 роки тому

      Squaring a function makes it differentiable. Hence.

  • @ArihantJain-h7x
    @ArihantJain-h7x 10 місяців тому

    Hi sir, thanks a lot for such valuable videos and crisp information.
    Can you please tell me why exactly a high coefficient value is a problem in regression models? Also is very low coefficient values also a problem?
    Thanks in advance.

  • @ilikasharma7957
    @ilikasharma7957 2 роки тому

    Thank you for informative video, how is accuracy less is in overfitting scenario?

  • @ArunKumar-pu9ko
    @ArunKumar-pu9ko 3 роки тому

    Which is the best/better regularization technique , and which is used for variable selection

  • @sujeshlx
    @sujeshlx Рік тому

    Great...

  • @nishah4058
    @nishah4058 11 місяців тому

    U told that l1 and l2 is only available for regression. But I have seen them for feature selection for textual dataset(although in textual data features are transformed into vector form and have numerical values) . So pls clarify the things that whether they used for feature selection also?

  • @iqbalbawa8875
    @iqbalbawa8875 2 роки тому

    Thank you for the explanation. Would have been useful to see how this would work in practice using an example in Excel using a small dataset or in Stata.

  • @abhishekkhairnar8318
    @abhishekkhairnar8318 2 роки тому

    nice explanation

  • @TheOraware
    @TheOraware 2 роки тому

    what do you mean by L1 and L2 regularization works only with Linear regression, decision tree based algo does have other way of regularization? You mean to say L1 and L2 are not in tree based algo? L1 and L2 are also used in decision tree based algo for example catboost regression has L2 (l2_leaf_reg) regularization technique

  • @sandipansarkar9211
    @sandipansarkar9211 3 роки тому +1

    finished watching

  • @sanyamsingh4907
    @sanyamsingh4907 3 роки тому

    Hi aman can I request you to make a video on what's the best approach of dealing with complex data in real world . as we know in real time the data is very unstructured and most of the time data doesn't exist in CSV form. But unfortunately many of the learning available on UA-cam is in perspective of analyzing data which is in CSV form . Can you please enlighten these points in your upcoming videos including the best and practical approach . For example how to work with JSON data in data science project ,, how to work with XML files etc?
    Regards
    Sanyam

  • @kshitizkhandelwal9348
    @kshitizkhandelwal9348 2 роки тому

    Sir since we already have learning rate to arrive at the optimum coefficients then why do we need to use Regularization ? Aren't both of them serving the same purpose?

  • @yogeshbabu3106
    @yogeshbabu3106 3 роки тому

    bro how do you find the equation (bo+b1) after find the fist cost function is high

  • @jolittevillaruz5234
    @jolittevillaruz5234 3 роки тому

    Is it possible to use ridge regression to impute univariate time series? Thanks

  • @Aaron_duckroast
    @Aaron_duckroast 3 роки тому +1

    Bro please include a exercise that uses all these.

  • @pinkakaur156
    @pinkakaur156 2 роки тому

    If my slope is coming very less but I want my model slope to be more what's thought in that

    • @UnfoldDataScience
      @UnfoldDataScience  2 роки тому

      Slope will come based on data, why u want to change it?

  • @beautyisinmind2163
    @beautyisinmind2163 2 роки тому

    sir can we use Lasso and ridge for feature selection in multi-class classification? say for IRIS data? or it is only for binary problem? please reply

  • @vishalreddyb9683
    @vishalreddyb9683 3 роки тому

    Sir, can you please make computer vision and CNN videos?

  • @shaelanderchauhan1963
    @shaelanderchauhan1963 3 роки тому

    I am not sure but we use l2 in neural networks i saw Andrew Ng lecture.

  • @prasannakumarthambharapu1449
    @prasannakumarthambharapu1449 2 роки тому

    sir, can you make more videos on deep learningg

  • @sagarvarandekar8279
    @sagarvarandekar8279 2 роки тому

    My lasso regression is getting wrong results. It is giving all coefficients as zero except the constant and R2 score as --0.001825328970232576. Someone please help.

  • @ravikumarranga8633
    @ravikumarranga8633 3 роки тому

    Linear Regression => (XTX)-1XTY
    Ridge Regression: (XTX+PI)-1XTY where P is penalty, I is Identity Matrix
    Lasso Regression=>please mention
    Elastic Net =>Please Mention

  • @mukeshkund4465
    @mukeshkund4465 3 роки тому

    Why L1 regularization creates sparsity ???

  • @eduardocasanova7301
    @eduardocasanova7301 3 роки тому

    Hi Aman, BIG FAN OF YOUR WORK!! I noticed you give DS training and filled the google form right away! Sadly didn't receive any email. Can you help me with my issue? Should i receive an email? I'm super interested.
    Thank you!

    • @UnfoldDataScience
      @UnfoldDataScience  3 роки тому +1

      Hi ,not getting enough bandwidth for training now, however I am working on a course, will share the update soon, many thanks for watching videos and staying connecetd.

    • @eduardocasanova7301
      @eduardocasanova7301 3 роки тому

      @@UnfoldDataScience Thanks for the quick response! Already turned on the notification bell!

  • @ayushtailor9941
    @ayushtailor9941 2 роки тому

    samjh nhi aaya bhai.where to use l1 and where to use l2? try explaining in dnn model

  • @Krishu_Patil2411
    @Krishu_Patil2411 3 роки тому +1

    Exact what i want AMAN

  • @Programmer92
    @Programmer92 24 дні тому

    will you help me sir

  • @rutviksavaliya2642
    @rutviksavaliya2642 3 роки тому

    not understood

  • @Prajwalsahuindore
    @Prajwalsahuindore 3 роки тому +1

    Bhai Hindustan me rhete ho toh hindi me bhi samjho na

    • @UnfoldDataScience
      @UnfoldDataScience  3 роки тому

      It's not about staying in India or anywhere else.
      Unfortunately we need to speak in English in office and interviews and this channel is completely in English so that everyone can understand.