# they’re doing it by degrees

Obviously it’s clear I’m a big nerd, but thinking back on the wonder that was Square One, I am astonished at how much work they put into their cultural references/parodies. You can dance if you want to… as long it is the Angle Dance!

On the official website you can watch the some other clips. If they ever released this show on DVD I would queue up to buy it.

# Postprint server for the California Digital Library

I’m surprised that there are fewer people using the Postprint server at the California Digital Libary (CDL). This is a service for University of California students, staff, and faculty that lets you post your paper in its final version and makes it freely available. The nice thing is that they pay attention to all the copyright details for you so you don’t have to sort that out. There are a few good reasons to use it : (a) it’s good to archive your work with your “employer,” (b) it helps promote open access to postprints for disciplines which don’t use ArXiV, and (c) they can give you statistics on number of downloads.

Just looking through it seems that a small fraction of the papers written by UC people make it into the repository. The University of California does not have a mandate for open access like Harvard does, but provides this as an option. As Michael Mitzenmacher has noted, mandates can become tricky, which is why the CDL’s managing of the copyright thing is nice.

# Is the minimizer a continuous function of the regularization parameter?

I am trying to figure out a rather simple question but I’m sure it’s known already. Suppose I have a the minimizer of the following regularized optimization:

$f^{\ast}(\lambda) = \arg \min_{f \in \mathcal{F}} L(f) + \lambda \|f\|^2$

Here $L(f)$ is a continuous convex function of $f$. Is $f^{\ast}(\lambda)$ a continuous function of $\lambda$? I want to say yes, but maybe I need more assumptions on $L(f)$.