12/2/11 UPDATE: Due to the evolving
nature of this course, and due to recent popular demand, ALL questions
highlighted in red will be EXTRA CREDIT on exam if everybody agrees during the
final review added on Wednesday 12/7/11. Maximum extra credit allowed will be
+30%. All other questions will be kept as candidate questions for Final.
=====================================================================================================================
The questions
for the final exam will BE DERIVED from the following (i.e., they may differ
slightly). The exam is closed everything. Also, since you have these questions
three weeks in advance, no questions regarding the exam will be answered prior
to the exam (from the time you receive this). Also, the point value of each
question will not be revealed until you receive the actual exam on test day.
EGR/CS
434
Green Robotics, Automation, and Machine
Intelligence
Final
Exam
Fall, 2011
Dr. Wunderlich
This is a closed-book, closed-notes exam (and
no calculators are allowed). Write your answers on the blank paper
supplied to you (write your name at the top of every page).
MOBILE
ROBOTS
IN REFERENCE TO THE FOLLOWING:
- PowerPoint: http://users.etown.edu/w/wunderjt/ITALY_2009/TALK_SOLAR_SYSTEM.pdf
- PowerPoint: http://users.etown.edu/w/wunderjt/ROME_2011_10.pdf
- PowerPoint: http://users.etown.edu/w/wunderjt/ITALY_2009/TALK_ROVERS.pdf
- PowerPoint: http://users.etown.edu/w/wunderjt/ITALY_2009/TALK_ROVER_MECHANICS.pdf
- PowerPoint: http://users.etown.edu/w/wunderjt/ITALY_2009/TALK_DELIVERY_SYSTEMS.pdf
- PowerPoint: http://users.etown.edu/w/wunderjt/ITALY_2009/TALK_POWER.pdf
- PowerPoint: http://users.etown.edu/w/wunderjt/ITALY_2009/TALK_SENSORS_&_NAVIGATION.pdf
- PowerPoint: http://users.etown.edu/w/wunderjt/ITALY_2009/TALK_AUVs_UUVs_and_Swarms.pdf
-
TALK: "Multi-Agent Robotics" (F12/9/11 by David
Coleman, Etown CENGR grad finishing U.Maryland Ph.D.)
and
READINGS:
- http://users.etown.edu/w/wunderjt/ROME_2011_PAPER9_434_Reading.pdf
- http://users.etown.edu/w/wunderjt/ITALY_2009/PUBLICATION_LRV.pdf
- http://users.etown.edu/w/wunderjt/ITALY_2009/PUBLICATION_LRV_Performance_Data.pdf
- http://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/20050217239_2005219510.pdf
-
http://users.etown.edu/w/wunderjt/ITALY_2009/PUBLICATION_MARS_SPIRIT_ALL_PATH_PLANNING_JPL.pdf
- http://users.etown.edu/w/wunderjt/ITALY_2009/PUBLICATION_MARS_SPIRIT_GLOBAL_PATH_PLANNING_JPL.pdf
- http://users.etown.edu/w/wunderjt/ITALY_2009/PUBLICATION_O3.pdf
- http://users.etown.edu/w/wunderjt/ITALY_2009/PUBLICATION_W0_COMM_Full%20Wunder2myrevision.pdf
- http://users.etown.edu/w/wunderjt/ITALY_2009/REPORT_James_Painter_w4_VISION_EGR494report.pdf
- http://users.etown.edu/w/wunderjt/ITALY_2009/REPORT_Jeremy_Crouse_w4_JAUS_EGR494report.pdf
- http://www.igvc.org/design/reports/dr207.pdf
- http://users.etown.edu/w/wunderjt/ITALY_2009/PUBLICATION_SrchRscBotsnSim_final2.pdf
- http://users.etown.edu/w/wunderjt/ITALY_2009/PUBLICATION14sim_vs_RT.pdf
- http://users.etown.edu/w/wunderjt/ITALY_2009/PUBLICATION_ASEEPAPetown2.pdf
- http://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/20050217239_2005219510.pdf
- http://users.etown.edu/w/wunderjt/ITALY_2009/PUBLICATION_MARS_SPIRIT_ALL_PATH_PLANNING_JPL.pdf
- http://users.etown.edu/w/wunderjt/ITALY_2009/REPORTandPHOTO_steveHendersonAQUABOT_SUBMITTED.pdf
- http://users.etown.edu/w/wunderjt/ITALY_2009/PUBLICATION_SrchRscBotsnSim_final2.pdf
- http://users.etown.edu/w/wunderjt/ITALY_2009/PUBLICATION14sim_vs_RT.pdf
- http://users.etown.edu/w/wunderjt/ITALY_2009/PUBLICATION_ASEEPAPetown2.pdf
- Course Text: Roland Siegwart, Illah R. Nourbakhsh and Davide Scaramuzza, “Introduction
to Autonomous Mobile Robots, Second Edition (Intelligent Robotics and
Autonomous Agents),” Massachusetts Institute of Technology, 2nd
edition, Mar 31, 2011.
ANSWER THE FOLLOWING:
#
(X points): Discuss the importance of “simplicity” in designing mobile robots
for autonomous missions (cite NASA examples)
#
(X points): Discuss the importance of “comprehensive quality testing” of
mobile robots for autonomous missions (cite NASA examples)
#
(X points): In Wunderlich, J.T. Designing robot autonomy: how
tightly should we hold the leash? The 5th International Conference on
Design Principles and Practices, Rome Italy, February 2011 (TALK, PAPER),
list all the major ways robot autonomy should be moderated and why.
# (X points): Concerning Mobile Robots (Siegwart
Text), name five fields of expertise
(in addition to some knowledge of kinematics, dynamics, and control theory),
that a mobile robot team should have.
# (X points): Concerning Mobile Robot Locomotion (Siegwart Text), name the three
reasons that replicating nature is
extremely difficult.
# (X points): Concerning Mobile Robot Locomotion (Siegwart Text), list for each of the key issues of “Stability,” “Ground-contact,” and
“Type-of-Environment,” two important issues.
# (X points): Concerning Mobile Robot Locomotion (Siegwart Text), list three advantages and two disadvantages
of legged robotic locomotion.
# (X points): Concerning Mobile Robot Locomotion (Siegwart Text), describe “Static Walking”
#
(X points): Concerning Mobile Robot Locomotion, describe why you would
ever want a mobile robot to have Humanoid
form.
#
(X points): Concerning Mobile Robot Locomotion (Siegwart
Text Chapter 2, and in ROVER MECHANICS TALK), Make a table with the following
column headings:
1.
“Maneuverability”
2.
“Controllability”
3.
“Stability(Static)”
4.
“Stability(Dynamic)”
5.
“Zero-Turning-Radius?”
6.
“Zero-Turning-Radius-in-it’s-footprint?”
7.
“Sketch of wheel
configuration with arrows drawn showing turning radius”
Then
compare a three wheeled robot of your choosing to both a standard
rear-wheel-drive car, and to the Wunderbot. Use “+” , “-“ , “~” to fill in table, and circle best in each
category.
# (X points): Concerning Mobile Robot Kinematics (Siegwart Text), list the five ways that a Mobile robot differs from a manipulator.
# (X points): Recalling
Siegwart Text and related PowerPoint,
describe in words, and vector diagrams
the relationship between a humanoid robot watching out the window his master’s
back yard which includes a mobile robot mowing the lawn using dead-reckoning
and random paths while avoiding a bunny who is chasing a butterfly. The
Humanoid must awaken his master if the lawn is done, if the mobile robot
accidentally traps the bunny, or if the bunny catches the butterfly; And tells
the master with a hand-drawing the exact location of the bunny, butterfly, and
the mobile robot. Clearly identify all local-to-global (and maybe
global-to-local) mappings. And clearly identify all simulations vs. real-time
control activities (both artificial and biological).
#
(X points): Describe how Dead
Reckoning works.
#
(X points): Describe how Celestial
Navigation works.
# (X points):
Concerning Mobile Robot Planning discuss with
pictures and words the methodology of path-planning
(LOCAL and GLOBAL) implemented in
all NASA Mar’s rovers (list all
rover names).
# (X points): Describe
all of recent sensors on the Wunderbots and discuss some of the complexities of the
sensor fusion.
# (X points): Discuss
how each of the following NASA rovers are powered: Moon LRV,
completed Mar’s Rovers, Future Mar’s and Europa Rovers
# (X points): In
David Coleman’s talk "Multi-Agent Robotics," discuss some of the details of his research
ROBOTIC
ARMS
IN REFERENCE TO THE FOLLOWING:
- PowerPoint: http://users.etown.edu/w/wunderjt/ITALY_2009/TALK_ARM_DESIGN.pdf
and
READINGS:
- Paper handout of Excerpts from Niku
Robotics Text
- Dr. W. 1993 lecture handout on Robot Kinematics
- http://users.etown.edu/w/wunderjt/ITALY_2009/PUBLICATION_SIM6_046338_Wunderlich_original_PROOF.pdf
- http://users.etown.edu/w/wunderjt/ITALY_2009/PUBLICATION_WeldingArm10_paperSUBMITrevisedFINAL.pdf
- Handout: Wunderlich, et. al. (1993). Software architecture for a kinematically dissimilar master-slave telerobot. In Proceedings of SPIE Int'l Conference on Telemanipulator Technology and Space Telerobotics, Boston,
MA: Vol. (2057). (pp. 187-198). SPIE Press.
ANSWER THE FOLLOWING:
# (X points): Concerning Robotic Manipulators (Niku Robotic Arms text handout), list any three advantages and any three disadvantages of Robots.
#
(X points): Concerning Robotic Manipulator DEGREES OF FREEDOM and referencing (Niku Robotic
Arms text handout), in Wunderlich, J.T. (2004). Simulating
a robotic arm in a box: redundant kinematics, path planning, and
rapid-prototyping for enclosed spaces. In Transactions of the Society for Modeling and Simulation International:
Vol. 80. (pp. 301-316). San Diego, CA: Sage Publications. (PAPER),
1.
How
many Degrees Of Freedom do you need to position the end of a manipulator
anywhere in 3D space
2.
How
many Degrees Of Freedom do you need to position and orient an
end-effector on the end of a manipulator anywhere in 3D space
3.
How
many Degrees Of Freedom does the human arm have.
Explain where they are and how they move.
4.
Name
SIX different things that can be optimized when you have “extra” degrees of
freedom (i.e., what can be done with the null-space)
# (X points): In Wunderlich, J.T. (2004). Simulating
a robotic arm in a box: redundant kinematics, path planning, and
rapid-prototyping for enclosed spaces. In Transactions of the Society for Modeling and Simulation International:
Vol. 80. (pp. 301-316). San Diego, CA: Sage Publications. (PAPER):
1. what is the main objective of the
research
2.
What obstacle avoidance technique is used
3.
How is the null space used
#
(X points): In Wunderlich, J.T. (2004).
Simulating a robotic arm in a box: redundant kinematics, path
planning, and rapid-prototyping for enclosed spaces. In Transactions of the Society for Modeling and
Simulation International: Vol. 80. (pp. 301-316). San Diego, CA: Sage
Publications. (PAPER)., describe in words the mathematical
complexity of attempting to control nine manipulator elbows simultaneously while
avoiding the walls of an enclosure while the end-effector performs a precise
task along the walls of a work cell deeply embedded in the enclosure. (All using a velocity-control scheme).
#
(X points): Concerning Robotic Manipulators (Niku
Robotic Arms text handout), describe with both pictures and words the three
different types of reference frames
used for manipulators.
#
(X points): Concerning Robotic Manipulators (Niku
Robotic Arms text handout,), describe two ways that a teach-pendant can be used by a factory technician to define the
operation of a manipulator.
# (X points): Concerning Robotic Manipulators (Niku Robotic Arms text handout), list 5 different types of industrial manufacturing applications discussed in lecture, and
five 5 different types of non-industrial
manufacturing applications discussed in lecture.
#
(X points): Recalling: Wunderlich, J.T. (2001).
Simulation vs. real-time control; with applications to
robotics and neural networks. In Proceedings of 2001 ASEE Annual Conference & Exposition,
Albuquerque, NM: (session 2793), [CD-ROM]. ASEE Publications, (PAPER):
1.
What
is the purpose of the Jacobian Matrix
2.
What
is the purpose of a unit-direction vector
#
(X points): In Wunderlich, J.T., S. Chen,
D. Pino, and T. Rahman
(1993). Software architecture for a kinematically
dissimilar master-slave telerobot. In Proceedings of SPIE Int'l Conference on Telemanipulator Technology and Space Telerobotics,
Boston, MA: Vol. (2057). (pp. 187-198). SPIE Press. (CLASS HARD-COPY HANDOUT ONLY):
1.
What
is the main objective of the research (i.e. who is it for and what will it do
for them)?
2.
What
is teleoperation?
AUTOMATION
IN REFERENCE TO THE FOLLOWING:
-
TALK: “Engineering in Impoverished Africa”
(10/2/11 by Dr. DeGoede)
-
TALK: “Phoenix Contact Industrial
Automation” (10/10/11 by Dan Fenton and Arnold Offner)
-
TALK: “Bechtel Engineering” (10/31/11
by Michael Mason, et. al.)
-
TALK: "Solar Charging Station
Grant" (11/17/11 by Offner,Vanderpool,&
Koep of Phoenix Contact)
-
TALK: "Etown
Sustainability Projects" (11/17/11 by Dr. Estrada & Dr.
Atwood)
-
TALK: "Siemens Building-Automation Products and Internships"
(W11/30/11 by Michael Baird, Siemens Branch Manager)
and
READINGS:
- PLC Book excerpts handout
- PLC Design example handout
ANSWER ONE OF THE FOLLOWING:
# (X points): What opportunities did Dr. DeGoede present as opportunities for
automation in Africa?
# (X points): What opportunities did Dan Fenton and Arnold Offner present as opportunities for
automation using Phoenix Contact technologies?
ANSWER ONE OF THE FOLLOWING:
# (X points): What opportunities did Michael Mason present as opportunities for Engineering at Bechtel?
# (X points): What opportunities did Offner,Vanderpool,&
Koep present
as opportunities for continuing the Solar Charging Station research?
ANSWER ONE OF THE FOLLOWING:
# (X points): What opportunities did Dr. Estrada & Dr. Atwood present as opportunities for Etown Sustainability Projects?
# (X points): What opportunities did Michael Baird present as opportunities for Siemens Building-Automation
Products and Internships?
# (X points): What is a PLC and how does it differ from a typical PC?
MACHINE
INTELLIGENCE
IN REFERENCE TO THE:
- PowerPoint: http://users.etown.edu/w/wunderjt/ITALY_2009/TALK_COMPUTERS.pdf
- PowerPoint: http://users.etown.edu/w/wunderjt/ITALY_2009/TALK_MACHINE_INTELLIGENCE.pdf
- PowerPoint: http://users.etown.edu/w/wunderjt/ITALY_2009/TALK_SENSORS_&_NAVIGATION.pdf
and READINGS:
- http://users.etown.edu/w/wunderjt/ITALY_2009/PUBLICATION_SUBMITcontrledRandSEcon.pdf
- http://users.etown.edu/w/wunderjt/ITALY_2009/PUBLICATION_CONTROLLED_RANDOMNESS_IBM.pdf
- http://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/20050217239_2005219510.pdf
- http://users.etown.edu/w/wunderjt/ITALY_2009/PUBLICATION_SrchRscBotsnSim_final2.pdf
- http://users.etown.edu/w/wunderjt/ITALY_2009/PUBLICATION14sim_vs_RT.pdf
- http://users.etown.edu/w/wunderjt/ITALY_2009/PUBLICATION_ASEEPAPetown2.pdf
- http://users.etown.edu/w/wunderjt/ITALY_2009/PUBLICATION_MARS_SPIRIT_ALL_PATH_PLANNING_JPL.pdf
- http://users.etown.edu/w/wunderjt/ITALY_2009/PUBLICATION_NNjournal19_SUBMIT_without_letter.pdf
- http://users.etown.edu/w/wunderjt/ITALY_2009/PUBLICATION_SUBMITPAPERdefininglimitsREVISED16b.pdf
- Neural network code handouts (Wunderlich, 1990-2011)
- Multivariable-calculus review and gradient
decent neural-network learning proof handout
- Rule-based AI theory and code handouts (Wunderlich, 1990-2011)
- Course Text: Roland Siegwart, Illah R. Nourbakhsh and Davide Scaramuzza, “Introduction
to Autonomous Mobile Robots, Second Edition (Intelligent Robotics and
Autonomous Agents),” Massachusetts Institute of Technology, 2nd
edition, Mar 31, 2011.
ANSWER THE FOLLOWING:
# (X points): Discuss the importance of simplicity
in designing computing systems for autonomous missions (cite NASA examples)
# (X points): Discuss the importance of comprehensive quality testing of computing systems for autonomous missions (cite
NASA examples)
#
(X points): Discuss some important differences in designing a real-time control system vs. a simulation
# (X points): Concerning Mobile Robot Perception (Siegwart Text), describe Proprioceptive and Exteroceptive sensing for mobile robots.
# (X points): Concerning Mobile Robot Perception (Siegwart
Text), describe Active and Passive
sensing for mobile robots.
#
(X points): Concerning Mobile Robot Perception
(Siegwart Text), list the seven classifications of
sensors.
# (X points): In Wunderlich, J.T. (2003). Defining the limits of machine
intelligence. In Proceedings of IEEE SoutheastCon, Ocho Rios, Jamaica, [CD-ROM]. IEEE Press.
: (PAPER),: how does symbolic AI differs
from artificial neural networks for each of the abilities listed here:
Reduced
Mental Ability
Matrix (i.e.,
reduce from the 42 abilities in paper)
|
|
|
Can human do? |
Can bug do? (spider) |
Can Conventional Computer Program do? |
Can Symbolic AI Program do? |
Can Artificial Neural Network do? |
Comments |
|
|
BASIC ANIMAL ABILITIES: |
|
|
|
|
|
|
|
3 |
Learn and adapt
|
yes |
yes |
no |
somewhat |
yes |
Evolution |
|
13 |
Generalize |
yes |
somewhat |
no |
somewhat |
yes |
|
|
14 |
Associate
|
yes |
yes |
somewhat |
somewhat |
yes |
|
|
15 |
Recognition patterns |
yes |
yes |
somewhat |
somewhat |
yes |
|
|
16 |
Robust under partial failure |
yes |
yes |
no |
no |
yes |
|
|
19 |
Stability, repeatability, predictability |
somewhat |
somewhat |
yes |
yes |
somewhat |
Uncertainty |
|
20 |
Multitask
|
yes |
yes |
yes |
no |
yes |
|
|
|
COMPLEX ABILITIES: |
|
|
|
|
|
|
|
21 |
Abstraction |
yes |
unlikely |
no |
no |
somewhat |
|
|
25 |
Heuristics |
yes |
yes |
somewhat |
yes |
no |
|
|
26 |
Inference |
yes |
yes |
somewhat |
yes |
somewhat |
|
|
27 |
Hypothesis testing |
yes |
somewhat |
somewhat |
yes |
no |
|
|
29 |
Ethical behavior |
yes |
unlikely |
no |
somewhat |
somewhat |
If coded/trained |
|
31 |
Open to inspection |
somewhat |
somewhat |
yes |
yes |
somewhat |
|
# (X points): In Wunderlich, J.T. (2004). Top-down
vs. bottom-up neurocomputer design. In Intelligent Engineering Systems through
Artificial Neural Networks, Proceedings of ANNIE 2004 International Conference,
St. Louis, MO. H. Dagli (Ed.): Vol. 14. (pp.
855-866). New York, NY: ASME Press: (PAPER), Briefly describe
the conceptual difference between "Top-down"
and "Bottom-up" neurocomputer design.
1.
Describe the importance of maintaining high-precision for the neuron
transfer function during learning.
2.
Compare and contrast Dr. W’s various numerical methods used to
approximate the neuron transfer function (just the basic concepts – don’t
reproduce all equations).
3.
Why bother to attempt on-chip learning?
4.
Why is a polynomial a good idea for neuron transfer function
implementation?
# (X points): In Campos, D. and Wunderlich, J. T. (2002). Development of an interactive simulation with real-time robots for
search and rescue. In Proceedings
of IEEE/ASME International conference on Flexible Automation, Hiroshima, Japan:
(session U-007). ASME Press: (PAPER), how are the Real-time robots
implemented and how do they communicate with the simulation?
#
(X points): List any five of the
10 parts of the
human brain identified in lecture
and discuss the function of each.
# (X points): What is the function of a dendrite.
# (X points): List
and describe the function of the 3 different types of neurons.
# (X points): Define
Proprioception.
# (X points): List the three
observations of the human brain made
by William James in 1890.
# (X points): Write Hebb’s 1949 statement
on learning.
#
(X points): In discussing all of the
historical developments that eventually led to the back-propagation neural network
model, briefly describe one
significant contribution of two of the
following individuals:
A.
1943
McCullah and Pitts
B.
1949
Hebb
C.
1958
Selfridge
D.
1960
Rosenblatt
E.
1974
Werbos
F.
1986
Rumelhart
#
(X points): Compare and contrast (with both sketches and words):
A.
Rosenbaltt’s 1960 PERCEPTRON
B.
Rumelhart et al. ‘s 1986 BACK-PROPAGATION
# (X points):
Concerning the research of the
Physicist J.J. Hopfield:
A. How did his 1982 neural network model’s learning differ from other models
such as backpropagation.
B. What did he contribute in 1984 that significantly changed the functioning
of other neural network models to follow. Explain the
significance of this contribution.
# (X points): Discuss a type of Neural Network different from
backpropagation and discuss how it differs.
# (X points): In
1969, Minsky and Papert
made a significant observation about neural networks regarding their ability to
deal with linear separability.
Define linear separability and compare using tables
and graphs the difference between the binary “AND” function and the binary
“XOR” function with regards to Minsky and Papert’s observations.
What development in neural networks solved this dilemma?
ANSWER ONE OF THE FOLLOWING:
# (X points): Concerning
neurons,
A.
Sketch
and label a graph of the action potential of a biological neuron
B.
Sketch
and label a graph of the hard-limiter
transfer-function used in artificial neural networks.
C.
Sketch
and label a graph of the sigmoid
transfer-function used in artificial neural networks.
D.
Which
of (b) or (c) is more like (a) and why.
E.
What
is it about the sigmoid transfer-function that makes it desirable for
back-propagation learning
# (X points): Concerning
artificial vs. biological Neural Networks,
A.
Sketch
and label a graph of the action potential of a biological neuron
B.
Describe
the basic functioning of the human brain and a biological neuron in your own
words (electrical and chemical concepts would enhance your answer, but are not
required).
C.
Sketch
and label a graph of the sigmoid transfer-function used in artificial neural
networks.
D.
Describe
the basic functioning of backpropagation neural
networks (learning and otherwise) in your own words.
# (X points): In gradient
decent learning, describe the import concept involved in:
A.
Varying the learning rate and how it relates to the topology of the
error-surface
B.
The importance of using a continuously differentiable neuron
transfer function and how the instantaneous-slope of this function can effect learning.
# (X points):
Using the Quotient Rule, find
the derivative of:
# (X points):
Find dy/dx for:
# (X points):
Describe in your own words and sketches (no equations), the method
of “Least Squares”
# (X points): Describe in your own words and sketches
(no equations), Gradient Descent
learning (including how the error surface is created)
# (X points): Sketch and label a graph of the typical sigmoidal neural network transfer function,
then sketch and label a graph of it’s derivative.
Then exp[lain how the magnitude of the derivative
effects neural network backpropagation learning.
# (X points): Given the following neuron transfer functions:
and error (i.e., “cost”) function:
derive using Calculus (i.e., the chain rule, partial
derivatives, the quotient rule, etc.) an equation for changing the weights
between the hidden layer and the output layer of a three-layered (i, j, k)
back-propagation neural network . Assume there are no BIAS connections to the
neurons. Simplify your final equation as much as possible. Also, discuss how
weight changes are magnified when output neurons are at “sticking points”.
# (X points):
Given the following neuron transfer
functions:
and error (i.e., “cost”) function:
derive using Calculus (i.e., the chain rule, partial
derivatives, the quotient rule, etc.) an equation for changing the weights
between the hidden layer and the output layer of a three-layered (i, j, k)
back-propagation neural network. Assume there are no BIAS connections to the
neurons. Simplify your final equation as much as possible.
# (X points):
Given the following neuron transfer
functions: . . . . .
and error (i.e., “cost”) function: . . . . .
derive using Calculus (i.e., the chain rule, partial
derivatives, the quotient rule, etc.) an equation for changing the weights
between the hidden layer and the output layer of a three-layered (i, j, k)
back-propagation neural network . Assume there are no BIAS connections to the
neurons. Simplify your final equation as much as possible. Also, discuss how
weight changes are magnified when output neurons are at “sticking points”.
# (X points): Why do you think much higher precision is needed for Neural Network back-propagation
computations during learning than
after learning completed.
#
(X points): Recalling Dr.
W.’s neural network code:
1.
What is the stopping tolerance for? (i.e., “STOPtolerance”)
2. Why do you think the learning rate can be set so high for simple
2-input examples (i.e. 1 to 5 instead of typically much less than 1 for complex
applications)?
3.
Why is the learning rate (in this code or any other
code) more sensitive for the XOR example than the AND or OR?
# (X points): In the matlab neural network toolbox example “Linear Design,”
what is the general concept of what this program is doing?
# (X points): In the matlab neural network toolbox example “Character
Recognition,” what is the general concept of what this program is doing?
How is noise
addressed?
#
(X points): Describe what each of the following historic Expert Systems does:
A.
DENDRAL
B.
MYCIN
C.
PROSPECTOR
D.
INTERNIST
# (X points): List and describe the
three major deficiencies of symbolic
A.I. expert systems
# (X points): List
and describe the four major types of
symbolic A.I. expert system problems
# (X points):
Describe “Backward-chaining” in
symbolic A.I. expert systems.
# (X points):
Describe what each of the following does in a symbolic A.I. expert system:
A.
Explanation Subsystem
B.
Knowledge Base
C.
Knowledge Base Editor
D.
Inference Engine
#
(X points): Draw a complete “And/Or
graph” for the following symbolic
A.I. expert system discussed in lecture defined by the following rules:
A.
If
engine is getting gas and the engine will turn over, then the problem is spark
plugs.
B.
If
engine is does not turn over and the lights do not come on, then the problem is
battery or cable.
C.
If
engine is does not turn over and the lights do come on, then the problem is
starter motor.
D.
If
there is gas in the fuel tank and gas in the carburetor, then the engine is
getting gas.
# (X points): Draw a complete “And/Or graph” for a symbolic A.I. expert system defined by
the following rules:
A.
If . . . . . , then . . . . .
B.
If . . . . . , then . . . . .
C.
If . . . . . , then . . . . .
D.
If . . . . . , then . . . . .
# (X points): Recalling Dr. W’s
symbolic A.I. expert system “Advise
Callers to a Doctor’s Office,” do:
A.
Describe
the important assumptions and possible ethical and legal ramifications of them.
B.
Do
one of the following (and state which one you are doing):
1.
Draw
the And/Or graph for Dr. W.’s design
2.
Create
you own rules, user input, etc. and draw your own And/Or graph
# (X points): Compare in your own words “Probability Theory”
vs. “Uncertainty Theory” for symbolic A.I. expert systems.
# (X points):
Concerning confidence values used
in Expert Systems:
A. List the equation for a “Certainty Factor,” and list the three laws
(equations) relating to it
B.
Using these equations, calculate the
following (show your work) . . . . . . .
# (X points): Describe the two places confidence values are used in Expert Systems
# (X points): Describe how Baye’s theorem
of Probability was applied to the “Copper Mining” problem discussed in
class.
# (X points):
Recalling Dr. W’s symbolic A.I.
expert system “Selecting a Toy for a Child,” do:
A.
Describe the significance of the Confidence values used for both the
rules and the user inputs.
B.
Draw the And/Or graph
DESIGN
IN REFERENCE TO YOUR TWO SEMESTER PROJECTS:
#
(10 points): Describe (in a Paragraph) something unexpected that you
discovered or realized as you worked on one of your projects.
#
(10 points): Describe (in a Paragraph) the most novel aspect of your
research or design.
#
(10 points): Describe (in a Paragraph) the most frustrating part of your
projects and how you overcame these hardships.