Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
11 Cards in this Set
- Front
- Back
|
How does connectionism define nodes? |
Nodes are processing units that contain minimal information. A single concept exists as a complex holistic pattern of activation involving very many nodes, distributed across several cortical areas. Like neurons, nodes can inhibit, excite, or activate proximal nodes in the network. The same set contains several concepts, depending on activation patterns. |
|
|
How do semantic-network models define nodes? |
Nodes in semantic networks bear information about a concept. The connection between two nodes is imagined as a labelled relationship; a proposition constitutes two nodes and their propositional relationship. |
|
|
How are spreading-activation processes explained in connectionism? |
Activation spreads throughout nodes in the network, to the limits of working-memory capacity. Nodes modulate proximal nodes; like neurons nodes can be excitatory or inhibitory, with several degrees of activation, and even partial activation. |
|
|
How are spreading-activation processes explained in semantic-network models? |
Nodes may indirectly activate neighboring nodes, with natural limits imposed by the weakening of activation farther from the initially active node. Proximal nodes receive stronger activation than distant nodes. Spreads easier along frequently active links than those seldom activated. |
|
|
How does connectionism fail to explain memories of single events, or single exposures to semantic information? |
Memories form as interconnected patterns of activated nodes, and forming new patterns happens when the same pattern of activation reoccurs more than once. In that model of memory formation, a single event shouldn't be capable of creating a well consolidated pattern of activation. |
|
|
How do connectionists propose to work around the issue of rapid unlearning, and memory consolidation after single events? |
McClelland et al. propose a complementary memory system that holds information for a short while, and with time is integrated with existing memory models, thus creating new activation patterns or moderating old patterns. |
|
|
What empiricism do we have for the 'second system workaround' in connectionism? |
Daily accumulation of 'temporary memories' appear to be integrated into proper long-term memory storage as we sleep. When awake, information enters the hippocampus from the neocortex through the entorhinal cortex; when asleep the flow reverses, and the neocortex is subjected to a flow of information originating in the hippocampus. |
|
|
In the ACT-R model, how is declarative information represented? |
ACT-R uses a semantic-network model based on propositions as the smallest unit of information you can judge the truth of. Connections between nodes include information about the relationship between subject node and object node. |
|
|
In the ACT-R model, how is proceduralisation modelled? |
Representation of procedural knowledge goes through three stages, this is known as proceduralization. The cognitive stage is an explicit deliberation of production rules. The associative stage is conscious practise of the production rules, in which we compose multiple production rules into more complex rules. The autonomous stage is using production rules automatically, and implicitly, without giving them conscious thought as we use them. |
|
|
How is the semantic-network model of declarative knowledge in ACT-R computational? |
A relationship between nodes is formal and binary; it is or isn't. Concepts are hierarchically categorised, and identifying a stimulus begins with the most salient feature and moves through sub-categories until the correct concept is activated. |
|
|
How is production-system representation of procedural knowledge in ACT-R computational? |
Procedures compose of linear sequences of operations, one at a time. Skills require serially applied production rules. Condition variables for "if-then" rules are explicitly binary, either-or. |
