Elements of guided-propagation (GP) | Definitions |
---|---|
GP architecture | Prewired set of processing modules defined by their relative connections within a matrix of channels and layers. The GP software deals with the modulation and growth from scratch of every module, thus simulating a global control device. Banks of elementary detectors and effectors form the predefined input/output of the full architecture |
GP-cells Elementary processing units (EPU) | Network nodes within every module. GP-cells are chained for encoding sequences of stimuli, giving rise to a dynamic tree-like structure that receives either stimuli or modulation from other modules. At least three types of GP-cells can be distinguished: the root unit, Context-Dependent units, and Event Detectors/Generators |
Context-dependent (CD) unit | Main GP-cell, possibly driven by a contextual input from its upstream predecessor along a tree branch. The other activating input is a stimulus generated by the output of another module. A CD unit responds to the possible coincidence between its Contextual (C) and Stimulus (S) inputs |
Event detector/generator (ED/G) | GP-cell located at the end of a module branch, among outputs which interface with other modules |
Memory path | Chained GP-cells, from a module root to an ED/G |
GP-cell excitability (E) and response threshold | The Excitability parameter Ei defines the total input required for cell i to fire, with a response threshold [S(t − τij) + C(t − τik)]/Ei, where τij codes for the time-delay of cell i input connection from cell j |
Ratio R between the two GP-cell inputs | Combined with E, the R parameter of a GP-cell allows its operating mode to be dynamically changed |
GP-cell operating modes | Different behaviors of a given cell (Fig. 4) are dynamically set by both computer program and running modules: |
1. Free | 1. Uncontrolled response (to be avoided) |
2. Inhibited | 2. No response |
3. Stimulus-driven | 3. Can respond irrelevant of the ongoing context (to be used with caution) |
4. Context-driven | 4. Can be fully activated by the only inner-flow (“Anticipation” mode) |
5. Restricted | 5. Requires both input to possibly respond (“quite”/Learning mode) |
Dynamic learning algorithm | A module set in the relevant (“Restricted”) operating mode allows the encoding of unexpected stimuli |
a. Differentiation | a. Sprouting of a new branch/’memory path’, provided free GP-cells. |
b. Generalization | b. Connection of a stimulus (ED/EG) with a memory path (CD unit). |