Lecture 10 | Convex Optimization I (Stanford)

[@shiv093]

1:15 Approximation and fitting 1:42 Norm Approximation 4:10 examples 5:39 Penalty function approximation 10:51 example 17:15 Huber penalty function 23:10 Least-norm problems 25:31 examples 30:17 Regularization approximation 34:53 Scalarized problem 39:03 Optimal input design 40:34 examples 42:48 Signal reconstruction 47:47 quadratic smoothing example 50:48 total variation reconstruction example 56:50 Robust approximation 1:06:19 stochastic robust LS 1:10:04 worst-case robust LS 1:15:52 example

[@ArnabJoardar]

Where are you, dude? Missing your bookmark keeping for all the remaining lectures.

[@shiv093]

@@ArnabJoardar My Professor skipped a few chapters. :(

[@tag_of_frank]

450,000 views for lecture one, 36,000 for lecture ten. About a factor of 10! Glad I made it through.

[@Gabend]

This is where the good stuff starts!

[@joshhyyym]

2020 and compressed sensing is still pretty popular in MRI

[@SalmaKhan-sy7tr]

Professor Boyd. The Great

[@ajminich]

Only Boyd could turn an annoying fly into a hilarious joke...and weave it into the entire lecture.

[@kylex7501]

32:50 A fly problem

[@ArnabJoardar]

At 8:46, he mentions that the 1-Norm has an opposite sense to that of the square-error function: for small residual errors, you're irritated more in case of the 1-Norm compared to the square-error function, relatively. Could anyone elaborate how a small residual value is able to give Large Irritation in the case of the 1-Norm? If the Irritant function was 1/residual, then, if the residual was small, i.e. 0

[@chaowu9105]

Arnab Joardar I think this is about the slope of L1 and quadratic function. L1 has constant 1 but quadratic has smaller than 1.

[@bgang6207]

"Large irritation" in the relative sense. Say the residual is 0.1, in 1-norm it is just 0.1, but in 2-norm it is 0.01 and 0.1>>0.01.

[@alialenezi360]

Don’t give up

[@TheProgrammer10]

the fly was just trying to get a better look at what boyd was talking about

[@annawilson3824]

1:10:00

[@jrf2623]

Lecture 10: optiflyzation problem

[@crystalp1983]

totally

[@EulerGauss13]

This feels like hearing all the toolkit without seeing and knowing anything about the real problems. Like I know this is applied part of the course, but it's so unmotivating.

[@何浩源-r2y]

Very funny