Today it’s raining in both Seattle and Paris. Here are a few things I think are worth reading while the weather clears up.
- Stop collecting coupons. Recently, Batson, Spielman, and Srivastava gave a beautiful sparsification result for graphs: Every graph has a linear-sized spectral sparsifier. Here is one statement of their result (taken from Srivastava’s thesis):
Theorem [BSS]: For every and , the following holds. For every , there are non-negative numbers of which at most are non-zero, and such that for all ,
Assaf Naor has given a very nice account of some recent geometric applications of this idea. These involves breaking the “coupon collecting” barrier from a number of results which were previously based on random sampling. There is still an outstanding open problem left open here: Embedding -dimensional subspaces of into .
- Lecture notes that are lecture notes. Some people write lecture notes like they are preliminary book drafts. These lecture notes by Ryan O’Donnell for an undergraduate course on “Probability and Computing” are like transcribed lectures. They’re conversational, with philosophical asides–a great example of the style.
- Metaphors in systolic geometry. Larry Guth and Nets Katz recently made awesome progress on the Erdos distance problem in the plane. On a different note, here is a great informal survey of Guth on Gromov’s systolic inequality (which itself answers the question—when would an isometric embedding into ever be employed for the usefulness of the ambient space?!)
- An important epsilon. Finally, there is the recent result of Gharan, Saberi, and Singh giving a approximation for the Traveling Salesman Problem in unweighted graphs. I haven’t gotten a chance to digest the paper yet. Here’s hoping someone else will write an overview of the key ideas.