mARC - Definition of memory Association by Reinforcements of Contexts

posted Jun 4, 2012, 11:19 AM by Jean-Michel Davault   [ updated Jun 26, 2012, 12:05 AM ]

When we present mARC, as a symposium or congress, we comment on a 3D short film showing how mARC works.

We present in this article :

- A definition of terms,

- An illustrative example, 

- The benefits of mARCTM .

 

Definitions:

An associative memory device is functionally equivalent to a traditional RAM or ROM. It stores binary data, regardless of the nature of these data.

The only difference is in how data is accessed :

·         In a RAM the binary data is accessible by an address representative of where the data is stored;

·         In an associative memory data is also accessible by address, but especially via data of the same nature as the data stored in memory.

memory Association by Reinforcements of Contexts”, a mARCTM goes beyond traditional “Associative memory” with “Enhanced Context” as mARCTM uses a method of detecting universal and binary contexts to organize, store and retrieve data.

context is a set of data connection connected to each others.

connection is a relationship of any kind between data, and implicitly carried by the flow of stored data.

Example: 
For text data, 
"Red Sea" is a close spatial relationship, 

"pressure bar" and "drink bar" associative relationships represent latent semantic type of language, 
"s" (plural) and " " (singular) is a language syntax relationship, 
"e" and "e" or "ê" express similarity relations between topological forms etc. ....

Benefits of a mARCTM

1 - Data Independence: In essence, a mARCTM acts as a signal processing device, independent of the nature thereof.

For example, a signal represented by a stream of text characters, mARCTMs contextual storage is language independent.

2 - Speed of Access: Context data speed of access is greater than, at least, one order of magnitude with respect to a RAM.

3 - Adaptativity: The input data can be recognized, even if they are noisy, provided contextual information allows. Partially altered information can be reconstituted.

4 - Compactness: A mARCTM self-regulates its space by factorising context internally, and ensures a representation as compact as possible of stored data. According to the nature of the implicit relationships of data, the representation provides maximum growth of (0) Log (n).

The more data are "naturally correlated”, in any way, the more the memory need will be low.

Typically, for textual data, the need for storage varies between (0) (n) at the beginning of storage, and (0) Log (n) at intermediate stage, to become asymptotically constant beyond a certain "critical mass".

5 - Easy programming: Once an information system needs a contextual data processing context, this process is fully supported by the mARCTM, simplifying the programming of such applications.


@ Soon