News that Matters?!?

Hello everyone!

I thought I'd inform all of you science geeks out there on two pretty cool advancements that were released yesterday.

The first, is the discovery of a compression algorithm that compresses "random" data up to a 100:1 ratio. Now, if you don't know a lot about compression algorithms (and if you call yourself a computer scientist you should!) then you probably wouldn't see the problem with this. The problem is that all current compression algorithms find patterns in the data to be compressed and use these repetitions to shrink the size of the data. The problems comes in when you consider Claude Shannon's Information Theory. In this theory (which is widely accepted), Shannon basically says that it is impossible to perform compression on "random" data. It is a fairly logical conclusion, cause if everything bit is important in the position it is in, how can you get rid of it and reduce the size of the information without lossing it forever? Well, supposidly ZeoSync has overcome information theory. There press release can be read: here

The second advancement deals with quantum computing (my fave!). It seems that scientists have found a way to store light in a crystal, and then re-emit that light from the crystal. This is a great advancement because it is basically a quantum hard drive. Though I wonder how well the crystal keeps the information held in the light away from outside interference (in order to make sure decoherence doesn't occur and mess up the calculations).

I thought these were two interesting advancements. Please feel free to give your opinions, or ask whatever questions you'd like, and I or another forum member would be more than happy to answer you.

Regards,
Wizeman

Wowser! ;)

There was a guy a while back (quite a while back) who had figured out how to use the light spectrum for memory (instead of this archaic toggling of bits, hehe). People in the US shunned the idea because the major products were still profitable so he went to Japan where they literally threw money at him for development. Wonder how it's going now...

But yeah, compressions 100:1 and using light/crystal for storage/speed? Hell yeah...bring it on!

Kool stuff! Good post Wizeman! :)

I know this is just a rehash ... but it seems boggling to me that you can have a random hash of data and have *any* significant compression. As far as I know, most established compression routines work somewhat like the following: every time it sees 001001001001001001001, for example, it replaces that with 'a' or something akin to that.

It seems that most compression routines use a cipher-esque hash and in this way crush data down.

All the same ... if this is true ... I wonder if the US will up its encryption standards?

Quantum hard drives are most definitely an interesting concept, however, I do not think it is new. Like Vorlin said, someone has tried this before. I remember hearing about this method for data storage some time back but never heard anything about it again. Perhaps it failed. Or perhaps Micorsoft bought out the technology and has been hiding it. Whatever happened, I hope this time we actually see some progress.

Quote:

---

Or perhaps Micorsoft bought out the technology and has been hiding it.

---

Actually, at least on the whole Quantum Computing thing, IBM has been the really big player. A couple months ago they were able to show, all by QC, that 5 and 3 are factors of 15 - which sounds stupid, but it's apparently the biggest QC computation to date, using 7 e- spins, if I remember correctly. Anyway - this story was /.'ed a while back (2 months? 3?)

~N~

Actually it was about a month ago. I posted about it on AO, but not a lot of people read it; if you are interested, go here . And actually, there has never been any sort of permanent storage for quantum computers, so this crystal is truly a first, but like I said I wonder how decoherence affects the information stored within the crystal.

Regards,
Wizeman

[QUOTE]Originally posted by nietzsche
I know this is just a rehash ... but it seems boggling to me that you can have a random hash of data and have *any* significant compression. As far as I know, most established compression routines work somewhat like the following: every time it sees 001001001001001001001, for example, it replaces that with 'a' or something akin to that.

-----------------------------------------------------------------------

Actually - 001001001001001001001 does not translate well - since binary bit representation is done it sets of 8 - but what the hell - this is how it would translate in ascii:

First Eight:
00100100 = 38 = &

The next Eight:

10010010 = 146 = ’

The next set:

Well - is only 5 bits - so I will assume you meant to put 3 ( or 00000011) zeros after it and it would be:

01001000 = 72 = H

So 00100100 10010010 01001000 would translate to " &’H " which is not really an " a " is it?

Or we could say that it represents "some number" - and then it would translate to:

001001001001001001001:
0+0+262144+0+0+32768+0+0+4096+0+0+512+0+0+64+0+0+8+0+0+1 = 299593
God - I must be bored if I just converted binary digits . . .

:)

------------------------------------------------------

If you guys really want to know how most compression algorithms do their encoding, then take a look at this. This describes Huffman Coding, the most prevalent encoding algorithm in use today.

Regards,
Wizeman

Quote:

---

Huffman Coding, the most prevalent encoding algorithm in use today

---

Mmmm ... how about Diffie or RSH? Since PGP is based (as far as my knowledge goes, which may well be dated, too) on RSH, I'd have thought this to be the most prevalent...?

Anyone have any figures on this? It'd be interesting, if a bit cerebral, to check out the stats. :)

Also, the problem (which seems to be implicitly covered in the document) is this: when you are building a state diagram, you have to have different "branches"... <backing off it a bit> okay - if you have a code which is "0011101", then you can't have a code which is "0" or which is "0011", as the FSM model won't know if you're going to put in the "011101" or "101" respectively ... so you solve this by either having a seperate control code at the start of each discrete element (i.e. "000111"), THEN you can do a "0001110" and a "0001110011", as it knows to keep going until it sees "000111", which is a new element. The other way to do this is with a tree, somewhat like the following:
0 1
0/\1 0/\1
('A')0/\1 ('C') ('x')0/\1

and so on.

This method seems to be covered in the article (shame on me, I just glanced through!).

Out of curiosity ... has anyone in here taken Automata Theory (upper-level Computer Science course)? If so, I'm sure that you could describe the FSM approach much better than I - I'm a bit rusty.

~N~

Um...I'm speaking of compression algorithms, it would seem you are talking about encryption algorithms...

Ummm ... yeah, you're right. Sorry about that!

Good catch. BTW: is Huffman used in LZH compression?

~N~