On: Card Trick Leads to New Bound on Data Compression ...

[colanderman]

What a surprisingly content-free article from MIT.

What bound did they break? By how much? In what domain does the bound apply?

(There is no one "compression bound" for lossless compression; there exist infinitely many lossless compression schemes which can compress at least one sequence of arbitrary length to a single bit, and every lossless compression scheme must have an expected compression ratio over all data of 1:1. Hence knowing the domain constraints is important!)

[judk → colanderman]

It's not surprising, it's technologyreview, which spun out from student newspaper to linkbaity web tabloid a few years ago.

In this case, however the original paper (already linked in the HN comments) is itself a linkbaity title with content-free abstract.

[3327 → colanderman]

Surprisingly "empty" article. "Well folks we've just lowered the bound on data compression - there you have it."

No reference, No Paper not even an example of how it applies. I hope an editor notices it and they pull it out - pretty embarrassing.

[grkvlt → 3327]

> Ref: arxiv.org/abs/1011.4609 Bounds from a Card Trick

At the bottom of the article

[kerkeslager → colanderman]

Yeah, I mean, if you ignore domain constraints, the following is a great compression algorithm:

    $infile = $ARGS[1]
    $outfile = $ARGS[2]
    touch $outfile

This was one of my favorite examples in Data Structures and Algorithms in college.

[colanderman → kerkeslager]

Except that's not lossless like I qualified.