Competitive Queuing Flashcards Preview

Neural Computation > Competitive Queuing > Flashcards

Flashcards in Competitive Queuing Deck (9)
Loading flashcards...
1
Q

What two assumptions are made in CQ?

A

(1) More than one plan representation can be simultaneously active in a planning layer; and
(2) The most active plan representation is chosen, in a second neural layer, by a competition run to decide which plan to enact next.

2
Q

What is activation in CQ and what is it used for?

A

“common currency” used to compare alternative plans

3
Q

How are choices made in CQ?

A

simple maximum-finding or WTA (winner-take-all) dynamics can be used as the choice mechanism in the choice layer

4
Q

What happens when a plan wins?

A

Once a plan wins the competition and is used to initiate a response, its representation is deleted from the field of competitors in the planning layer, and the competition is re-run.

5
Q

How does CQ allow for a serial performance and who?

A

This iteration allows the two-layer network to transform an initial activity distribution across plan representations, often called a primacy gradient, into a serial performance (Grossberg, 1978).

6
Q

What is present in the current state?

A

All the information needed to specify a forthcoming sequence is present in the current state of the planning level of a CQ system, although this current state may itself be dynamically evolving.

7
Q

How would CQ be used to output a sequence of letters?

A
  1. The most active plan is selected for execution in the lower, competitive choice, layer by a winner-take all dynamic whose outcome is wholly determined (in the absence of noise) by the activation gradient (representing the to-be-performed sequence) present in the parallel planning layer.
  2. Once a plan representation wins at the competitive layer, a large output signal is sent to initiate execution of the corresponding response (descending arrow) and to delete the plan’s representation in the parallel activation layer (ascending path to parallel planning layer).
  3. This process iterates until all plans have been enacted and all planning layer activities deleted
8
Q

What is the goal of CQ?

A

sequential plan execution that corresponds to the initial rank ordering (primacy gradient) of plan activation levels in the upper field of the CQ network

9
Q

How does CQ use inhibition?

A

each competitive layer node would send an inhibitory connection to its correspondent in the parallel planning layer

recurrent inhibition can bee used in the choice layer, each competitive layer node would inhibit all others, but only nearest-neighbor inhibition is actually depicted