Sunday, May 1, 2016

Self adapting statistics and the queuing model, a simple example

Examine the grocery store, where the manager's goal is to balance customer flow and employee flow.  At the check out stand, one or two customers in line is optimum.  Then what iss his the optimum queue for clerks?  There will always be at least one clerk who is not busy and tends to front counter inventory.  The store manager is self adapting, managing an exit rate (clerk queue) with the entrance rate (customer queue).  Thew theory of everything tells us that when the queues are balanced, then one complete, typical sequence of transactions will occur before.  The store manager has encoded the typical transaction sequence the drives his business.  That sequence should be one complete set of unique school girls walking abreast on a Sunday afternoon.

There should be a decomposition theorem, building one complete sequence from others sub-sequences, ask a pro.  I look at it as the decoding/encoding graph, which is self adapting, an adaptive Huffman encoder.

Currency banking

The currency banker is willing to risk new currency in service of price discovery.  The currency banker wants a lower and uer bou d on this float.  It basically runs an adaptive Huffman encoder on member banks deposits / loans balance.  The precision is pricing is pre-announced, so if the precision is 3%, the currency banker promises to generate then decoding tree that  will generate a unique sequence of 32 deposit/loan balance for its member banker, and each member bank will be charged, or receive payment to move their banances into the quantized range.  

So, one or more member banks find a great investment opportunity, and they generate consistent deposit and loan transactions.  Their significance causes the adaptive decoding tree to alter shape, the tree must have 32 leaves. Each member bank wins or loses depending upon how close they are to the new typical sequence. The connection to queuing is the decoding network, it has to be an optimum queued network.

In this system, when more significant inside information is uncovered, less important information is discarded.  This is unlike block chain which supports infinite precision.  The marginal gains por loses are quantization noise.  The mathematicians get this.

No comments: