Machine Learning Links and Boooks

Here I will post good links on things related to ML, statistics and math.

Books

• Christoph Molnar's book on interpretable ML looks very good: https://christophm.github.io/interpretable-ml-book/index.html

There are big courses, and then there are BIG courses

Having TA'd large lecture courses in grad school, I know how much work goes into teaching them.  But this is an entirely different scale:

“Stanford University has opened up to the public an introductory artificial intelligence class, taught by two luminaries in the field: 70,000 people have signed up to receive registration information for the course (blogs.scientificamerican.com)” 

70,000 is a lot of assignments to grade.  I feel feel sorry for the TA’s.

Bill O'Reilly is Such an Embarassment

Bill O'Reilly's gall, hubris and ignorance are astounding:

• http://www.youtube.com/watch?v=3XEgkViLbTk
• http://www.youtube.com/watch?v=UyHzhtARf8M
  

Fox has to get rid of him. Who cares how much money he brings in. If they had any standards, they'd let him go. Stephen Colbert put it best (after the Anderson Cooper bit). O'Reilly's understanding of the world really is from the 13th century.

More on NP-Incomplete Problems (Ladner's Theorem)

I did a bit of searching regarding my previous post, and it is known that if P != NP then there do exist NP Incomplete problems. They're actually called "NP Intermediate" rather than "NP Incomplete" problems. I like my name "NP Incomplete" better but the term "complete" may have a specific meaning that doesn't apply to the NP intermediate class --- I don't know.

The existence of NP incomplete problems was proved (assuming P != NP) by Richard Ladner in 1975. I haven't looked at it but hope to soon. Some well known problems thought to be NP Intermediate are:

1. Graph Isomorphism
2. Factoring

Garey and Johnson also mentions Linear Programming as a candidate for NP intermediate, but we now know that LP's are in P.  It's good to remember we are making progress.

NP Incomplete Problems?

In the wake of Vinay Deolalikar's maybe proof that P != NP, I found myself thinking about the old P ?= NP question. I realized that while I know of a many problems that are NP complete, I don't know of any problems that are:

1. in NP
2. not known to be in P
3. not known to be NP-complete

Are there such "NP incomplete" problems? And has anyone proved that if P not equal to NP, such NP-incomplete problems must exist? It would be surprising if they didn't since the NP-complete problems are the hardest problems in NP. Without an NP-incomplete class, there would be a sudden jump from so-called "easy" (in P) to hardest (NP-complete).

NBC's "Law and Order" receives ultimate punishment

There may no longer be "Law and Order" in the world, but at least there's justice. Belinda Goldsmith reports that the long-running criminal justice drama has finally been canceled. Unfortunately, we will still be victimized by it's various spin-offs, as they are slated to continue.

A Supreme Court without Protestants?

Headline from today's Boston Globe on the nomination of Elena Kagan: "If she’s confirmed, court would have no Protestants." Who is this relevant to in 2010? I kind of thought we were past this stuff when we elected a black man for president...