Imagine you are a painter with odd paints in three cans. Each can occupies a region on a three color graph (red,yellow,blue),but the tint keep changing bit in each can.
Your job, as an artist, it to sequentially select daps of paint from each of the cans and fill in the 'numbers' along a counting index space. Your job is done when, on stepping back the sequence looks white, the colors optimally blend so as to eliminate tint and bound the brightness.
The artist is oloroit boss. It is creating a generating queue structure that captures all the redundancy in the dabs of paint along the sequence, leaving only the unexpected variations. Bit error becomes uni form white noise.
It is a simple model, and applies from quarks and gluons to central banking and debt cartels.
This simple index model is sort of general purpose, as it does not presuppose any physics algebra. The Shannon model of information tetory, as flow, tellus a property on the indexed output.
The theory says the index space can be subdivided, as needed, into a twos bit ruler such that the granularity of the ruler is sufficient to simulate the artist stepping back and looking. It is the data geek down the hall who is told t find N bit fixed point processor that is sufficient to the task. Hence the idea of adaptive means asymptotically stable, the adapter always points bit error back to white. The slope of adaption can be modelled as a N bit fraction.
The sandbox cares about Shannon because of the Walmart shopper. Our shopper carries with him an N bit computer, his cash card and budget manager. and acts like a one color balance sheet. The checkout stand looks like a static noise wait in line, he does not need a model of supply, the clerks. The singularity is going to do twos bit math, but the act of shopping is a one color experience, the shopper can assume fixed rices while walking the aisles, and fixed queue structure on checking out. This is the original Shannon.
No comments:
Post a Comment