Okay, so what we’re going to do is we’re basically going to replace this indicator function with an approximation of the indicator function that’s smooth. So and this is being many ways to talk – a lot of people talk about this. It’s something like this. When I’m making fun of people who propose things like this I usually say this is the same as getting out some sandpaper and saying you don’t like this corner here so you sand off the corner or something like that. ...

See the whole transcript at Convex Optimization I - Lecture 18