PSY30400 Chapter Notes - Chapter 9: Connectionism
Chapter 9: Knowledge
• How are concepts represented in connectionist network?
o Model using connectionist ideas as proposed by McClelland and Rogers
▪ representation of ord anary is represented y pattern of atiity in all
units in the network
▪ A anary is a… is represented in netork y atiation of property units
plus pattern of atiation of netork’s representation and hidden units
• However, a connectionist network MUST BE TRAINED in order for this
result to occur
o Training inoles adjusting netork’s CONNECTION WEIGHTS
• Example:
o If the model is UNTRAINED and all connection weights are 1.0,
then when presenting canary and can, all words with weight
1.0 will be activated, such as daisy and tall
o If model is to work properly, then connection weights must be
adjusted so that activating concept unit canary and relation
unit can only activates property units of grow, more, fly, and
sing
• Learning process occurs when erroneous responses in property units
cause error signal to be sent back through network by back
propagation
find more resources at oneclass.com
find more resources at oneclass.com
