Hi, Why is it easy to optimize via coordinate ascent to solve dual optimization?
Could you explain more about Coordinate descent?
I have seen the last part of lecture note, It is not clear enough, like what is g(alpha)?
Also a question from Lab 7, what is dual objective? I have no idea of the exact form of it.
The dual problem is equivalent to the minimization of a convex quadratic function over a decomposable convex set. Therefore it has a specific structure which is suitable to coordinate descent. This is precisely the object of the second exercice of the lab 7.
Your question about coordinate descent is a little unclear and I invite you to come tomorrow to the Q&A session if you want more explanations about this algorithm.
To reply to your question, \(g(.)\) is any function you want to minimize using coordinate descent and \(\alpha\) is the variable you are optimizing on.
With respect to the Lab 7, I do not see where the dual objective is referred to but the dual optimization problem is precisely defined in Equation (2).
Do not hesitate to use the exercice sessions to ask your TA more details on both the lecture and the lab and to come to the Q&A session.
Question on Coordinate descent
Hi, Why is it easy to optimize via coordinate ascent to solve dual optimization?
Could you explain more about Coordinate descent?
I have seen the last part of lecture note, It is not clear enough, like what is g(alpha)?
Also a question from Lab 7, what is dual objective? I have no idea of the exact form of it.
The dual problem is equivalent to the minimization of a convex quadratic function over a decomposable convex set. Therefore it has a specific structure which is suitable to coordinate descent. This is precisely the object of the second exercice of the lab 7.
Your question about coordinate descent is a little unclear and I invite you to come tomorrow to the Q&A session if you want more explanations about this algorithm.
To reply to your question, \(g(.)\) is any function you want to minimize using coordinate descent and \(\alpha\) is the variable you are optimizing on.
With respect to the Lab 7, I do not see where the dual objective is referred to but the dual optimization problem is precisely defined in Equation (2).
Do not hesitate to use the exercice sessions to ask your TA more details on both the lecture and the lab and to come to the Q&A session.
Best,
Nicolas
Add comment