Cog Sci Exam 3 UPDATED Exam
Questions and CORRECT Answers
Physical Symbol System Hypothesis - CORRECT ANSWER - Symbols represent
knowledge that are manipulated through thinking and searched through intelligence
Multiply realizable - CORRECT ANSWER - Symbols in PSSH can exist in various brain
forms.
Thinking - CORRECT ANSWER - Manipulation of symbols into new symbols.
Search space - CORRECT ANSWER - States and rules that describe all path combinations
to get possible solutions for a given problem. A type of intelligence
General Problem Solver by Newell and Simon - CORRECT ANSWER - AI program
simulating human problem-solving.
Heuristic search - CORRECT ANSWER - Problem-solving method by only considering
options that are immediately available
River Crossing Problem - CORRECT ANSWER - An example of defining a search space
and then completing it via heuristic search
Who is Shakey - CORRECT ANSWER - Early robot demonstrating physical symbol
system.
What can Shakey do - CORRECT ANSWER - navigate and identify obstacles in a space
Connectionist network - CORRECT ANSWER - Model mimicking brain's neural
connections. Each unit is not actual neurons.
, Weighted connections - CORRECT ANSWER - Strength of links between connectionist
units.
Feed-forward network - CORRECT ANSWER - Information flows in one direction only.
Hidden units - CORRECT ANSWER - Intermediate processing units in neural networks.
How are connectionist networks programed - CORRECT ANSWER - The number of
states and connections are set up with random values for the connections. Then data is sent
through the model and it's connections are modified using back propagation
Back propagation algorithm - CORRECT ANSWER - Method for training neural
networks. Takes result of running a datapoint through the model and modifies the connection
strengths.
Comparisons of Connectionist networks and the human brain - CORRECT ANSWER -
Units are not neurons, brain is much bigger, the general function is the same
Local optimal solution - CORRECT ANSWER - Best solution within a limited area.
Global optimal solution - CORRECT ANSWER - Best overall solution across entire
search space.
How to always get a global optimal solution - CORRECT ANSWER - Run back
propagation with many different starting values for the weighted connections
Overfitting - CORRECT ANSWER - Model too closely matches training data so it cannot
adjust to new test data.
Questions and CORRECT Answers
Physical Symbol System Hypothesis - CORRECT ANSWER - Symbols represent
knowledge that are manipulated through thinking and searched through intelligence
Multiply realizable - CORRECT ANSWER - Symbols in PSSH can exist in various brain
forms.
Thinking - CORRECT ANSWER - Manipulation of symbols into new symbols.
Search space - CORRECT ANSWER - States and rules that describe all path combinations
to get possible solutions for a given problem. A type of intelligence
General Problem Solver by Newell and Simon - CORRECT ANSWER - AI program
simulating human problem-solving.
Heuristic search - CORRECT ANSWER - Problem-solving method by only considering
options that are immediately available
River Crossing Problem - CORRECT ANSWER - An example of defining a search space
and then completing it via heuristic search
Who is Shakey - CORRECT ANSWER - Early robot demonstrating physical symbol
system.
What can Shakey do - CORRECT ANSWER - navigate and identify obstacles in a space
Connectionist network - CORRECT ANSWER - Model mimicking brain's neural
connections. Each unit is not actual neurons.
, Weighted connections - CORRECT ANSWER - Strength of links between connectionist
units.
Feed-forward network - CORRECT ANSWER - Information flows in one direction only.
Hidden units - CORRECT ANSWER - Intermediate processing units in neural networks.
How are connectionist networks programed - CORRECT ANSWER - The number of
states and connections are set up with random values for the connections. Then data is sent
through the model and it's connections are modified using back propagation
Back propagation algorithm - CORRECT ANSWER - Method for training neural
networks. Takes result of running a datapoint through the model and modifies the connection
strengths.
Comparisons of Connectionist networks and the human brain - CORRECT ANSWER -
Units are not neurons, brain is much bigger, the general function is the same
Local optimal solution - CORRECT ANSWER - Best solution within a limited area.
Global optimal solution - CORRECT ANSWER - Best overall solution across entire
search space.
How to always get a global optimal solution - CORRECT ANSWER - Run back
propagation with many different starting values for the weighted connections
Overfitting - CORRECT ANSWER - Model too closely matches training data so it cannot
adjust to new test data.
