Information Technology Reference
In-Depth Information
Second, in the well-learned stages of typing (skilled typist in [17]), when stim-
uli to be typed are repetitive letters (e.g., AAA...), M1 is strongly activated, how-
ever, when stimuli to be typed are multiletter sentences (e.g., JACK AND...), M1 is
strongly activated, but there is more robust activation in the SMA, the basal ganglia,
and S1.
9.2.3 A Queuing Network Model with Reinforcement Learning
Algorithms
9.2.3.1 The Static Portion of the Queuing Network Model
Queuing network is a mathematical discipline that is used to simulate and model a
wide array of phenomena and systems including manufacturing and computer net-
work performance. A queuing network is a network of servers that provide services
to customers that wait in queues before they are serviced. Queuing networks tend
to be quite flexible and can allow two or more servers to act in serial, in parallel,
or in any network configuration [26, 27]. Computational models based on queuing
networks have successfully integrated a large number of mathematical models of
response time [26] and multitask performance [27]. A queuing network modeling
architecture is called the queuing network. Model human processor (QN-MHP) has
been developed and used to generate behavior in real time [28], including simple and
choice reaction time [14] and driver performance [44]. The model in this chapter ex-
tends QN-MHP by integrating reinforcement learning algorithms and strengthening
its long-term memory and nine motor subnetwork servers. In addition, the queuing
network approach has also been used to quantify changes in brain activation for
different participant populations [4].
The brain, which is an enormously complex network of interconnected systems
and subsystems, acts in concert with one another to produce behavior. This idea is
supported by evidence from pathway tracing studies in nonhuman primates, which
revealed widely distributed networks of interconnected cortical areas, providing an
anatomical substrate for large-scale parallel processing in the cerebral cortex [6]. It
seems, then, that brain areas do not act in isolation from another and instead may
form complex neural networks that are the basis of behavior and thought.
In addition to the widely distributed nature of the brain, each brain area may also
have some level of functional specialization [9] and thus each major brain area may
have certain information processing capacities and certain processing time parame-
ters (see Table 9.1). Here we assume that the interconnections between major brain
areas form a queuing network with each major brain area composing a queuing
network server and that information processed at each server is a queuing network
entity. In addition, neuron pathways that connect major brain areas serve as routes
between our queuing network servers (see Fig. 9.1 for transcription typing routing
and Fig. 3.1 a for PRP routing. Note that both networks have the same servers and
overall network configurations). Therefore, it is assumed that the major brain areas
form a queuing network with brain areas as the servers, information processed as
