Chronology Current Month Current Thread Current Date
[Year List] [Month List (current year)] [Date Index] [Thread Index] [Thread Prev] [Thread Next] [Date Prev] [Date Next]

Re: [Phys-l] iPod music compression



Folkerts, Timothy J wrote:
I just had a random thought as I was getting ready to help students
review for their next test: loading info to your brain is a little like
loading songs to an iPod!

If you simply try to memorize every bit (literally!) of information in
raw, uncompressed form, you will end up with very little actually
information. Only a few uncompressed songs will fit on an iPod!

On the other hand, compression looks for patterns in the data so that
less memory is required to reproduce the same end result. If the
students invest a bit of time to learn the patterns - the theories,
models, equations - then they can "compress" other raw information so
that their brains will hold much more useful information overall!
Is this a silly analogy, or is there enough truth that it might help
students think about how to study & learn?

Like all analogies, this one is inexact, but it is far from silly.

The issue with human memory is not so much getting stuff to "fit";
the issue is being able to recall it when needed. That involves
exploiting the connections between ideas.

As a related point, it is easy to learn ideas that are consistent
with previously-learned ideas. That is, it helps if the new
ideas are at least partially predictable based on previous ideas.

Lossless compression depends on predictability. Music is partially
predictable; the next millisecond is partially predictable based on
what went before. This is intimately related to ideas of information
and entropy.

Conversely, the whole point of learning is to make predictions. If
the material to be learned was not predictable, there would be no
point in learning it.

That means there is an if-and-only-if relationship between compression
and learning. (This can be quantified quite precisely, but the
details are kinda complicated.)

Learning also involves being able to separate what's important from
what's not important. Lossy compression exploits some observed universal
properties of human perception.

Typical coders use a combination of lossless and lossy coding methods.