*Thanks. Step added! (cstheory)*

(Please ignore the previous post; it has some errors)

I read your blog and found it curious. Thought that I too will go through the expander survey.

However, I find it confusing that while demonstrating the existence of super-concentrator, you first use the case where you have k n/2 edges, you rely upon the matching between L1 and L2 as identified in the survey.

I mean, if we wanted to use that matching why do not we use it for k < n/2 edges as well?

Also I fail to understand why Valiant "ever" conjectured that any super-concentrator must have "much more than" O(n) edges?

Consider the simple bi-partite graph with n-vertices on left and n on right. Each L(i) is joined to R(i). Sure, this graph is disconnected but I guess the definition allows it.

Please answer if you have time (and patience!) to reply to my (possibly stupid!) queries

]]>I read your blog and found it curious. Thought that I too will go through the expander survey.

However, I find it confusing that while demonstrating the existence of superconcentrator, you first use the case where you have k n/2?

I mean, if we wanted to use that matching why do not we use it for k > n edges?

Please answer if you have time (and patience!) to reply to my (possibly stupid!) queries

]]>The blogs I list do discuss important stuff but I still miss detailed technical posts. ]]>

Also, the difference between CS theory and math these days is vanishing…

]]>