Friday, November 4, 2016

Missing the obvious

I think we should have known, look at the Huffman encoding tree. It only works if symbols do not bunch up on that tree. A little proof using conservation of symbols and standard queueing theory does the trick.  Shannon is a theory of queueing on a network. Then we generalized that to self adapting queue networks, getting us the hyperbolic model, and a whole bunch of stuff I forgot already..

Helps with design.  A bin full of bid/ask sitting on a node is a queue, it is likely five wide, a long queue.    That is quantization noise, bids and ask annihilates, and their variance in the bin has to be allocated, round off error is like radiated energy.  There is flow, the bubble up, eventually forces annihilation.  Flow, queue rebalancing, queuing; say with that.  Then we are not confused when physicists talk about information in subatomic thought.

Instead, physicists should say, things get bunch up, the vacuum gets compressed, or there will be a gap in the queue,empty space, egad; or if they bunch, say :"look, the bubbles start to loop and spiral into not-space when bunched".! Stuff like that, if the engineers want to know about time and distance, give them an approximate derivation.

Some analogies
What happens in a re-quant when the bit error gets split evenly? You get mass, encapsulated light.
When events bunch up? You either violate maximum entropy, your bin sizes up the graph ate not at equal variance. (Arrivals between nodes is not a guassian bet, hence the arrival pattern will have self induced phase interaction. A redundancy, looping, not maximum entropy.
When you get that Gibbs lock? When the clock energy goes to relative one, the hyperbolic condition met, the quantization well deepens.  In this state,there is no combination of arrivals that results in fewer transitions. The 'clock' drops to a minimum as mass and light split the residual. There is not enough red/green risk taking, you get an atomic orbital.


No comments: