What a beautiful lecture it is, thanks Prof. Dutta. It's a great introduction on Convex Optimization. The lecture takes you from the very basic of defining a global/local minimum of a function to deriving the necessary conditions for a function to have a minimum evaluated at a certain point. The Mathematics although rigorous but have been explained in a most logical and lucid terms so that you fall in love with it. The best part of this lecture was how it sets up the platform to explore questions like What is Convex optimization and what's the need of that and the boundary of this course. Thanks a lot.
Dear Abhilash. The function at 3: 55 is indeed continuous and is nice since it is piecewise linear. I would be happy if you let me know what are the vague statements you have found. For an optimizer it is the discontinuous functions that are not nice because there really no good ways to handle them. Here the function is piecewise linear and continuous and you have algorithms to find an approximation to the local minimum.
Dear Sir , @44 .03 o||λw|| = λ , although implicitly understood , could you please prove it as you said the norm of the vector multiplied with order is the order itself . Sorry , not explicitly clear on that part. Else it's an excellent explanation overall
Dear sir I am a big fan of your lectures.but I have a suggestion.All the IIT maths videos are very abstract in nature.after sometimes we feel that we are doing something among symbols , but so intense mathematical session is not good for engineers , because I think they forget about physical realities at all. while if you see standford convex optimization or Alex smola phd level machine learning videos or even a very simple economics optimization videos available on youtube , you will find that they are smart to integrate both things.I am sorry if you feel it is not a good idea.I advise you to delete the comment in that case.
Not really!! This kind of rigor is absolutely imperative for doing good research regardless of whether one is a mathematician or an engineer. For context, I am an Electronics Engineer and would've been very disappointed with less rigorous exposition.
I found the lecture excellent. It conveyed intuition very nicely, without being muddled by details from applications. This is math conveyed in an agreeable manner.
The Stanford course on convex optimisation was “awful”. Massive diversions during the course, not explaining certain things properly at all, and half the lecture going away arguing semantics which make no sense! This is my experience with the first 5 lectures at Stanford on convex optimisation.
Sure it is vague but not wrongs. It is humanly impossible to avoid vague descriptions completely in lectures. But generally they themselves declare that it is a vague description.
Of course it is a continuous function and not differentiable on those sharp edges. In a MOOC setup you can't precisely say everything, although in this lecture the details are immaculate. Somebody wrote in a comment that it is an aggreable mathematics between a teacher and a student. Cheers !!
What a beautiful lecture it is, thanks Prof. Dutta. It's a great introduction on Convex Optimization. The lecture takes you from the very basic of defining a global/local minimum of a function to deriving the necessary conditions for a function to have a minimum evaluated at a certain point. The Mathematics although rigorous but have been explained in a most logical and lucid terms so that you fall in love with it. The best part of this lecture was how it sets up the platform to explore questions like What is Convex optimization and what's the need of that and the boundary of this course. Thanks a lot.
Loved this series, helped me a lot while studying optimisation in machine learning
Dear Abhilash. The function at 3: 55 is indeed continuous and is nice since it is piecewise linear. I would be happy if you let me know what are the vague statements you have found. For an optimizer it is the discontinuous functions that are not nice because there really no good ways to handle them. Here the function is piecewise linear and continuous and you have algorithms to find an approximation to the local minimum.
Dear Professor, Could you please let me know what are the prerequisites/linear algebra topics I should learn before starting this course
Playlist link: ruclips.net/p/PLbMVogVj5nJQHFqfiSdgaLCCWvDcm1W4l
I enjoyed your lecture. Thanks for the thorough explanation
Great lecture... Explained everything in detail
Dear Sir , @44 .03 o||λw|| = λ , although implicitly understood , could you please prove it as you said the norm of the vector multiplied with order is the order itself . Sorry , not explicitly clear on that part. Else it's an excellent explanation overall
It is a very good course.
The one by @ahmadbazzi is better
Sir,
Can you point to any Textbooks as a recommendation to supplement the lectures?
Thank You
Here is the Course Syllabus
nptel.ac.in/content/syllabus_pdf/111104068.pdf
Nice lecture sir
Dear sir
I am a big fan of your lectures.but I have a suggestion.All the IIT maths videos are very abstract in nature.after sometimes we feel that we are doing something among symbols , but so intense mathematical session is not good for engineers , because I think they forget about physical realities at all.
while if you see standford convex optimization or Alex smola phd level machine learning videos or even a very simple economics optimization videos available on youtube , you will find that they are smart to integrate both things.I am sorry if you feel it is not a good idea.I advise you to delete the comment in that case.
Not really!! This kind of rigor is absolutely imperative for doing good research regardless of whether one is a mathematician or an engineer. For context, I am an Electronics Engineer and would've been very disappointed with less rigorous exposition.
I found the lecture excellent. It conveyed intuition very nicely, without being muddled by details from applications. This is math conveyed in an agreeable manner.
The Stanford course on convex optimisation was “awful”. Massive diversions during the course, not explaining certain things properly at all, and half the lecture going away arguing semantics which make no sense! This is my experience with the first 5 lectures at Stanford on convex optimisation.
@47:41 it is not necessary condition e.g take y = |x|.
He is talking about differentiable functions only.
@3:55 the function is nice and continuous? lectures must stop using vague and wrongs descriptions!
Sure it is vague but not wrongs. It is humanly impossible to avoid vague descriptions completely in lectures. But generally they themselves declare that it is a vague description.
Not sure what you mean. The function is continuous though it's derivative is discontinuous.
Of course it is a continuous function and not differentiable on those sharp edges. In a MOOC setup you can't precisely say everything, although in this lecture the details are immaculate. Somebody wrote in a comment that it is an aggreable mathematics between a teacher and a student. Cheers !!