Awesome video, Professor. I no longer attend UW-Madison, i went grad school elsewhere, but I always come back to these videos. Thanks for all your help during 2018-2020!
Thanks for the kind words, Arhum. Glad you got something useful out of these classes, and wishing you all the best for your current studies and future career!
I think it's worth mentioning that even if the features at 3:12 overlap, it only appears that way based on the feature scaling shown. When the red x's and blue o's are normalized, the blue circles could in fact have greater discriminatory power than features which are separated farther apart. And as you said, without the labels there is not way of you knowing if those features are even useful or not, they could just be uniformly distributed over a greater range.
Sir I am applying various feature engineering technqiues for my problme (Phishing detection) it's a multiclass problem. I am using huge dataset (25 input features) and 600000 rows. What techqniues I should prefer more. Waiting for your answer. Thanks
Prof. I have big confusion so I visited here. Can we apply embedded lasso and ridge regularization feature selection technique for the dataset having multiclass classification problem? for example dataset like IRIS???please give me authentic answer with some reference if possible
Yes, sure, that's not an issue at all. In the Lasso lecture we actually use a dataset with multiple classes. The same is true for the sequential feature selection lecture etc.
![Megan Thee Stallion - Bigger In Texas [Official Video]](http://i.ytimg.com/vi/QxTkZTLWdo4/mqdefault.jpg)
![Summer Walker - Heart Of A Woman [Official Lyric Video]](http://i.ytimg.com/vi/SFddctbZzw8/mqdefault.jpg)
Awesome video, Professor. I no longer attend UW-Madison, i went grad school elsewhere, but I always come back to these videos. Thanks for all your help during 2018-2020!
Thanks for the kind words, Arhum. Glad you got something useful out of these classes, and wishing you all the best for your current studies and future career!
@@SebastianRaschka Thanks Professor. I saw that you’ll be on leave starting next semester. Will you be moving to NYC as you start at grid.ai?
@@arhumz1573 It's a remote position, so I don't know about moving, yet, but I will surely be visiting NYC a lot in the future :)
I think it's worth mentioning that even if the features at 3:12 overlap, it only appears that way based on the feature scaling shown. When the red x's and blue o's are normalized, the blue circles could in fact have greater discriminatory power than features which are separated farther apart. And as you said, without the labels there is not way of you knowing if those features are even useful or not, they could just be uniformly distributed over a greater range.
Sir I am applying various feature engineering technqiues for my problme (Phishing detection) it's a multiclass problem. I am using huge dataset (25 input features) and 600000 rows. What techqniues I should prefer more. Waiting for your answer. Thanks
Prof. I have big confusion so I visited here. Can we apply embedded lasso and ridge regularization feature selection technique for the dataset having multiclass classification problem? for example dataset like IRIS???please give me authentic answer with some reference if possible
Yes, sure, that's not an issue at all. In the Lasso lecture we actually use a dataset with multiple classes. The same is true for the sequential feature selection lecture etc.
@@SebastianRaschka Thank you professor, I'm glad to receive your reply