RMSprop Optimizer Explained in Detail | Deep Learning

[@kasyapdharanikota8570]

your channel is highly underrated, it deserves a lot more audience

[@MachineLearningWithJay]

Thank you for this considerate comment 😇

[@RajeswariThiyagarajan-d3k]

thankyou so much for uploading these videos, your explanations are easily understandable

[@vishalchandra1450]

Hand's down,best explanation ever:)

[@MachineLearningWithJay]

Haha… Thanks you so much 😄

[@himtyagi9740]

waiting for SVM since you explain so nicely..thnks

[@MachineLearningWithJay]

Thank you! I will upload SVM video after finishing RNN series

[@Vinay192]

Hi Sir, any plan of uploading videos on support vector machines? If yes then, please try to cover the mathematical background of SVM as much as you can ... Anyway your content is really appreciable...Thanks !

[@MachineLearningWithJay]

Thank you so much for your suggestion! Yes, I will be making video on SVM and covering mathematical details behind it.

[@KorobkaAl]

You are the best, thanks dude 🤙

[@MachineLearningWithJay]

You’re welcome 😇

[@jyotsanaj1425]

Such clear explanation

[@Maciek17PL]

If situation with w and b would be opposite values of gradients on the vertical axis were small and values on horizontal axis where large would RMSprop slow down the training by making vertical axis values larger and horizontal axis values smaller?

[@MachineLearningWithJay]

No no… it will still make the training faster. Vertical horizontal is just an example i am giving. Realistically, it can be in any direction. In every direction, its gonna work the same way.

[@ueslijtety]

Hi，Is it correct that you set the vertical coordinates to w and the horizontal coordinates to b? I think it should be the other way around.Because whether the goal can be reached in the end depends on w rather than b.

[@MachineLearningWithJay]

Hi… neither we set vertical to w nor b. Its just an example given… in a model.. there are many axis, not just x and y if we have more than 2 number of features. So a model can take any axis as any w or b. and it doesn’t matter as well which axis is for waht

[@ueslijtety]

@@MachineLearningWithJay thanks!So in practice this is not going to be a 2D planar image but a multidimensional image?And which parameters can determine the point of convergence in gradient descent?W OR b?

[@minister1005]

So i guess what he means is that if you get a high gradient, you will be updated a lower amount and if you get a low gradient, you will be updated a higher amount.

[@marccasals6366]

You're incredible

[@MachineLearningWithJay]

Thank You Marc! Glad you found my videos valuable.

[@christopherwashington9448]

Hello thanks for the info. But you didn't mention the purpose of the square for the gradient.

[@MrMadmaggot]

Man and what kind of LOSS should I use when training using RMSprop optimizer?

[@MachineLearningWithJay]

You can use any loss function

[@keshavmaheshwari521]

What is S?

[@syedalimoajiz1179]

how to initialize value of Sdw and Sdb?

[@yahavx]

What is (dw)^2?

[@JyothiPulikanti-o6m]

Explain ADMM also

[@mugomuiruri2313]

good

[@Garg478]

“THUS” is not pronounceda as “THAS” it’s “DAS”