Comments for I’m a bandit
https://blogs.princeton.edu/imabandit
Random topics on optimization, probability, and statistics. By Sébastien BubeckFri, 11 Sep 2015 03:18:41 +0000hourly1Comment on A solution to bandit convex optimization by Sebastien Bubeck
https://blogs.princeton.edu/imabandit/2015/07/24/a-solution-to-bandit-convex-optimization/comment-page-1/#comment-347240
Fri, 11 Sep 2015 03:18:41 +0000http://blogs.princeton.edu/imabandit/?p=990#comment-347240Haha :), I don’t even know if this qualifies as a galactic algorithm: in some sense we are just giving an analysis of the optimal minimax algorithm (which can be approximated in time T^T, just use multiplicative weights on a discretization of the space of all algorithms for this problem…).
]]>Comment on ORF523: Nesterov’s Accelerated Gradient Descent by Sebastien Bubeck
https://blogs.princeton.edu/imabandit/2013/04/01/acceleratedgradientdescent/comment-page-1/#comment-347237
Fri, 11 Sep 2015 03:13:18 +0000http://blogs.princeton.edu/imabandit/?p=386#comment-347237These are very good questions, both answered by Nesterov. For accelerating Newton’s method see this paper: http://link.springer.com/article/10.1007%2Fs10107-006-0089-x ; and for accelerating coordinate descent see this: http://www.optimization-online.org/DB_FILE/2010/01/2527.pdf .
]]>Comment on ORF523: Nesterov’s Accelerated Gradient Descent by Coordinate Descet
https://blogs.princeton.edu/imabandit/2013/04/01/acceleratedgradientdescent/comment-page-1/#comment-346612
Tue, 08 Sep 2015 02:28:26 +0000http://blogs.princeton.edu/imabandit/?p=386#comment-346612Is it possible to apply the nesterov acceleration to the second order newton method? and to the block coordinate descent method?
]]>Comment on A solution to bandit convex optimization by Igor Carron
https://blogs.princeton.edu/imabandit/2015/07/24/a-solution-to-bandit-convex-optimization/comment-page-1/#comment-346446
Mon, 07 Sep 2015 08:10:09 +0000http://blogs.princeton.edu/imabandit/?p=990#comment-346446So you now have your own galactic algorithm ( https://rjlipton.wordpress.com/2010/10/23/galactic-algorithms/ )
]]>Comment on Deep stuff about deep learning? by DEEP LEARNING REVOLUTION, summer 2015, state of the art & topnotch links | Turing Machine
https://blogs.princeton.edu/imabandit/2015/03/20/deep-stuff-about-deep-learning/comment-page-1/#comment-345546
Tue, 01 Sep 2015 21:29:34 +0000http://blogs.princeton.edu/imabandit/?p=957#comment-345546[…] 5. Deep stuff about deep learning? | I’m a bandit […]
]]>Comment on Welcome! (with random graphs) by Abdullah
https://blogs.princeton.edu/imabandit/2013/01/13/rgg/comment-page-1/#comment-341156
Wed, 29 Jul 2015 17:56:55 +0000https://blogs.princeton.edu/imabandit/?p=11#comment-341156In the first line of re-writing P(Y_k=0) with Markov inequality.

How is P(Y=0) *<=* P((Y-EY)^2=(EY)^2), would *=* be more precise ?

]]>Comment on ORF523: Mirror Descent, part II/II by Eufrasio
https://blogs.princeton.edu/imabandit/2013/04/18/orf523-mirror-descent-part-iiii/comment-page-1/#comment-339720
Thu, 16 Jul 2015 15:49:08 +0000http://blogs.princeton.edu/imabandit/?p=560#comment-339720Congratulations for your post. Do you know where can I find a r code (or library – http://www.r-project.org) with the mirror descent algorithm?
]]>Comment on Nesterov’s Accelerated Gradient Descent for Smooth and Strongly Convex Optimization by [Unfinished] Gradient-based optimization algorithms for training in machine learning and EDA global placement process | Exponentials
https://blogs.princeton.edu/imabandit/2014/03/06/nesterovs-accelerated-gradient-descent-for-smooth-and-strongly-convex-optimization/comment-page-1/#comment-339513
Tue, 14 Jul 2015 21:27:29 +0000http://blogs.princeton.edu/imabandit/?p=814#comment-339513[…] [2] Sebastien Bubeck, Nesterov’s Accelerated Gradient Descent for Smooth and Strongly Convex Optimization, https://blogs.princeton.edu/imabandit/2014/03/06/nesterovs-accelerated-gradient-descent-for-smooth-a… […]
]]>Comment on Revisiting Nesterov’s Acceleration by Anonymous
https://blogs.princeton.edu/imabandit/2015/06/30/revisiting-nesterovs-acceleration/comment-page-1/#comment-338076
Thu, 02 Jul 2015 09:59:25 +0000http://blogs.princeton.edu/imabandit/?p=983#comment-338076Another relevant development is a recent paper by Kim and Fessler that reaches a method very similar to AGD via a fully constructive approach.
It doesn’t give an answer to “why”, but at least gives a non-mysterious “how”.
]]>Comment on COLT 2015 accepted papers and some cool videos by Annabelle Smyth
https://blogs.princeton.edu/imabandit/2015/05/04/colt-2015-accepted-papers-and-some-cool-videos/comment-page-1/#comment-329622
Thu, 21 May 2015 22:43:10 +0000http://blogs.princeton.edu/imabandit/?p=965#comment-329622It’s amazing to think about the universe and the different potential dimension of it. Most of this goes over my head, but the things I do understand are fun to think about!
]]>