PARALLEL DISTRIBUTED PROCESSING (PDP)
| Collection | International Encyclopedia of Systems and Cybernetics |
|---|---|
| Year | 2004 |
| Vol. (num.) | 2(2) |
| ID | ◀ 2468 ▶ |
| Object type | Discipline oriented, General information, Human sciences |
Template:Ency entity are softwares for use in parallel computers. They emulate neural networks in the brain and are used to “train” nets in these computers to construct representations of specific situations
RUMELHART enumerates the major aspects of a parallel distributed processing model as follows:
a set of processing units (“neurons ”)
- a state of activation y$_{i}$for every unit, that determines the output of the unit
- connections between theunits . Each connection is defined by a weight w$_{ij}$, which determines the effect of thesignal of unit i on unit j
- a propagation rule ( a governing equation) , which determines the effectiveinput u$_{i}$ of a unit from its external inputs
- an activation function F which determines the new level of activation based on the effective output u$_{i}$ (t) and the current state y$_{i}$(t)
- an external input or offset $_{i}$for each unit
- a method for information gathering
- an environment within which the system must operate, providing input signals and — if necessary- error signals
(J. Mc CLELLAND & D. RUMELHART: “Parallel Distributed Processing Explorations in the Microstructure of Cognition” 1986).
This is the basic connectionist interpretation of the brain as the massively parallel computer, which depends on the general and local architecture of the connections between elements and on the dynamics of synaptic weights adjustments.
R. FISCHER explains Template:Ency entity, applied both to artificial and natural intelligence, as follows; “In this… parallel distributed processing approach, learning takes place through changes in the system itself” (1992, p.208).
This comment seems to establish limits to the organizational closure concept: Before closure, organization must be constructed (but probably on the base of organizational closure already acquired at a lesser level of complexity, biological at least).
FISCHER adds: “The principle of Template:Ency entity implies that activities of ordered sets of nerve cells can be considered to be mathematical vectors. An important aspect of this vector approach is that it focuses on the explanation of brain functions in terms of neural networks, i.e. mass action and that it is, therefore, compatible with the modular organization of the brain… The complexity of the neural circuitry of the cerebral cortex can be looked upon as a neural correlate of intelligence” (p.221).
FISCHER even observes: “At the interface of human mind functions and A.I. lurks the thought that the mind is trying to create intelligent systems in its own image” (p.226).
M. BODEN applied the Template:Ency entity models to the activity of a group of children: “The class-decision is due to the parallel processing (all the children chatter simultaneously) of localized computation (each child speaks to, and is directly influenced by, only her immediate neighbours) and is distributed across the whole community (as an internally consistent set of mini-decisions made by all the children).
In Template:Ency entity models, concepts are represented as activity-patterns across a group of units (Children)“ (1990, p.121).