Applied artificial intelligence resources
Applied behavioral robotics : neurosemantic networks.
Neurosemantic networks (NSN) are auto-organized, behavior-driven neural networks. The general purpose of an NSN is to assemble basic motor units, which consist of elementary behavior patterns, in order to produce new computer objects, after a learning phase. To understand how they work, we could compare them with the behavior of a group of fishes in the water : thanks to its sensors, every single fish is aware of its current environment (let us say the state of the environment at time t). Thanks to its motor organs, the fish will also be able to (re)act to the state of the environment, which means to modify the latter. This first reaction will modify its own future perception of the same environment at time t+1, which will lead to a new reaction, and a new change to the environment. And so on. In addition, the fish will not behave the same way if alone in the water or if living within a group of similar fishes : it is due to the fact that every other fish will also (re)act and modify the common environment, thus the perception of the latter by every other fish at time t+1. In this scheme, a neurosemantic network represents the group of fishes, the output object being the environment. The size of the network determines the predictibility of the future states of the environment, at times t+1, t+2, t+n.
Using smaller network size result in higher predictability, and still manageable chaos.
Unlike neural networks (which use and produce values), neurosemantic networks only use computer objects as input, and produce other objects as output.
Fields: neural networks, NSN's, pattern recognition, sensors, motor organs, analyzers, generators, semantic analysis, artificial intelligence, behavioral robotics, chaos modelization, auto-organizing networks, kohonen nets, low predictibility automats, multicellular behavior, organic life, logical robots, artificial life, low level deterministic systems.