The Streamer uses group theory (instead of sHe's category theory [More]) to achieve massive scalability over diverse sensor types and recognition modules. An implementation using IBM’s remarkable InfoSphere Streams architecture [More ] is underway.
The human brain is a multipurpose organism that supports diverse reasoning and memory functions. Computer systems designers often use the brain as a rough analogy, hence the term “artificial intelligence.” In the human, the brain is fed by a network of senses, and the nerves that carry that information actually have a subtle intelligence of their own. In between the eye and brain is the optic nerve which provides a sort of in-stream pre-recognition by sophisticated pattern identification..
The Streamer multiplies this “eidetic” system to industrial proportions. What if you had a million eyes — in fact a million instances of thousands of senses — and you wanted the associated nerves to collaborate, so they could present pre-reasoned structures to the brain? We have many such sensors, but they cannot yet exchange information. or even report common abstractions. Some of these media collectors are human-inspired (such as video cameras) while others are inscrutable in their raw form (such as synthetic aperture radar). The problem of many stream types, yet no way to have in-stream collaborative recognition exists in the military intelligence community as well as many business domains.
Topoiesis [More] is the logical framework that allows the synthesis to happen. It is a two-sorted logic where the second “sort” assembles information about the present situation based on prior knowledge and the evolution of the situation thus far. Where more heavy, unbounded reasoning is the focus of the intelligence workstation [More], and virtual enterprise framework [More], the streamer focuses on computational efficiency, just-in-time analyses and massive scalability.
Where the other engineering solutions work on sophisticated topological structures in category theory, the streamer uses more computationally cheap group theoretic products.
The work of Michael Leyton [More ] and his wreath products can be used.