Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
35 Cards in this Set
- Front
- Back
|
What is Fuzzy Logic?
|
Fuzzy logic is a form of many-valued logic that deals with approximate, rather than fixed and exact reasoning. Compared to traditional binary logic (where variables may take on true or false values), fuzzy logic variables may have a truth value that ranges in degree between 0 and 1.
|
|
|
What is Bi-valent logic?
|
Black or white, true or false, 0 or 1.
|
|
|
What is Multi-valent logic?
|
Shades of grey, values between 0 and 1.
|
|
|
4 main stages of Case Based Reasoning? |
Retrieve - the most similar cases after comparing current to the library of past cases. Re-use - reuse the retrieved case to try and solve the current problem. Revise - revise and adapt the proposed solution if required. Retain - retain the final solution as part of a new case. |
|
|
Knowledge can come in 2 forms, what are they. |
Knowledge can be simple or complex. |
|
|
Main roles of a knowledge engineer during the process of knowledge acquisition.
|
Assess the problem.
Elicit the knowledge. Structure the knowledge. Validate the knowledge. |
|
|
What is a reactive agent? Also advantages and disadvantages.
|
A reactive agent is is an agent that is focused on fast reaction to changes detected in the environment. E.g. sensor in front of robot, robot then changes movement left/right when sensor detects obstactle.
|
|
|
What is a utility function?
|
●Maps a set of states to a set of real numbers. E.g. Given a particular state of the world, an agent is able to use it's utility function to derive a score (utility value) that tells it how 'happy' it is in that state or how successful it has been if it reaches that state. ●Best applied to whole run as this provides us with insight into current state with a goal oriented approach. |
|
|
What are preference heuristics? Describe 2 and how they might be triggered in an application.
|
Specificity preference - used for conflict resolution. If 2 rules apply to a situation, but one has a more specific condition, then one will be chosen in preference to the other, other things being equal. Ease-of-adaption preference - |
|
|
What is an Agent? (with reference to those studied in class)
|
An agent is anything that can be viewed as perceiving it's environment through sensors and acting upon that environment through effectors. |
|
|
How do expert systems attempt to manage uncertainty? |
Uncertainty arises from from values that occurs between true and false. Add a belief measure to indicate the extent to which the conclusion is true. Attach numerical value an uncertain/ purely qualitative outcome. E.g. if condition then action (0.85). Expert developer considers recommended action can ONLY be trusted with HIGH level but not TOTAL certainty. These certainty factors help to develop systems that are more responsive to complex environments. |
|
|
What is Simple Knowledge? |
Traffic light sequence etc. |
|
|
Example of Complex Knowledge? |
Expert domain knowledge, e.g. brain surgeon, pro chess player. |
|
|
What is a GA? (define it in 5 steps) |
· Generate a random population of chromosomes · If the termination criteria are satisfied, stop, else go to step 3 · Evaluate this population using some ‘fitness’ criteria · Apply crossover and mutation operators to select the next generation of chromosomes · Return to step 2 |
|
|
When to use CBR? |
● when a domain model is difficult orimpossible to elicit ● when the system will require constantmaintenance ● when records of previously successfulsolutions exist |
|
|
CBR v Rule Based systems. |
● CBR offers a cost-effective solution to the ‘knowledge acquisition bottleneck’problem ● CBR systems can learn from experience and so can be self-maintaining ● Rule-based systems are better when itis hard to gather case data |
|
|
What is subsumption? |
●To design an agent as a set of task accomplishing behaviours, arranged into a subsumption hierarchy. ●Each task behaviour implemented as a finite state machine, which directly maps sensor input to effector output. ●Layers interact with each other via suppression and inhibition actions. E.g. a lower layer in hierarchy, representing low level behaviour (such as avoiding obstacles) may suppress higher layers that represent more abstract behaviours (such as exploring or avoiding obstacles). The process of designing an agent becomes one of systematically adding and experimenting with behaviours. |
|
|
Most KRL's made up of 2 interlinked features. What are they? |
Most KRL's represent data and phenomena in terms of objects and relationships between objects. KRL's influence by graph theory (such as semantic net), objects denoted by nodes or vertices, and relationships by links or edge. |
|
|
What is a Proactive agent? |
Pro-active / goal-directed: Agents are often deemed goal-directed, having goals toachieve (not necessarily objectives to maximise) with respect to their behaviours’.For example, agents within a geographic space can be developed to find or follow aset of spatial paths to achieve a goal within a certain constraint (e.g. time), whenexiting a building during an emergency. |
|
|
Goal of planning agent? |
• Choose actions to achieve a certain goal
• then use logic to represent: – Actions – States – Goals |
|
|
What does STRIPS action schema do? |
STRIPS is a planning language that represents plan components as states, goals, and actions. The action schema consists of 3 parts: • action name and parameter • pre-condition - a conjunction of function free positive literals stating what must be true in a state before the action can be executed • a conjunction of function free literals describing how the state changes when action is executed. |
|
|
About Abstraction, essential points. |
•Captures essential features of a domain, working as a means to managing complexity. (removes noise and redundant data) • Decisions have to be made between efficiency and expressiveness. •Up to developer to question what may need to be retained or what can be ignored. |
|
|
2 abstraction examples in expert systems. |
Fuzzy systems use domain abstraction, objects in a domain are grouped into equivalence classes. Semantic networks involves grouping nodes preserving some property. |
|
|
What is knowledge? |
Knowledge is a theoretical or practical understanding of a subject or a domain. Knowledge is also the sum of what is currently known, and apparently knowl-edge is power. Those who possess knowledge are called experts. They are themost powerful and important people in their organisations. Any successfulcompany has at least a few first-class experts and it cannot remain in businesswithout them.
Anyone can be considered a domain expert if he or she has deep knowledge (ofboth facts and rules) and strong practical experience in a particular domain. Thearea of the domain may be limited. For example, experts in electrical machinesmay have only general knowledge about transformers, while experts in lifeinsurance marketing might have limited understanding of a real estate insurancepolicy. In general, an expert is a skilful person who can do things other peoplecannot. |
|
|
What is a knowledge expert? |
Anyone can be considered a domain expert if he or she has deep knowledge (of both facts and rules) and strong practical experience in a particular domain. The area of the domain may be limited. For example, experts in electrical machines may have only general knowledge about transformers, while experts in life insurance marketing might have limited understanding of a real estate insurance policy. In general, an expert is a skilful person who can do things other people cannot. |
|
|
What is intelligence? |
The ability to learn from, under, and interact with one's environment. Intelligent people recognise their strengths and weakness, then figure out how to capitalist on their strengths, which compensates for their weaknesses. Intelligent people are able to find a balance between abilities such as analytical, creative, and practical. |
|
|
What is intelligent machine behaviour? |
. A machine is thought intelligent if it can achieve human-level performance insome cognitive task. To build an intelligent machine, we have to capture,organise and use human expert knowledge in some problem area. |
|
|
Simulated Annealing definition: |
Each Each set of a solution represents a different internal energy of the system. Heating the system results in a relaxation of the acceptance criteria of the samples taken from the search space. As the system is cooled, the acceptance criteria of samples is narrowed to focus on improving movements. Once the system has cooled, the configuration will represent a sample at or close to a global optimum. |
|
|
GA Advantages: |
It always gives solution and solution gets betterwith time.
It supports multi-objective optimization. It is more useful and efficient when search spaceis large, complex and poorly known or no mathematical analysis is available. The GA is well suited to and has beenextensively applied to solve complex designoptimization problems because it can handleboth discrete and continuous variables, and nonlinear objective functions. without requiringgradient information. |
|
|
GA limitations: |
When fitness function is not properly defined,GA may converge towards local optima Operation on dynamic sets is difficult. GA is not appropriate choice for constraint based optimization problem. |
|
|
SA advantages: |
It statistically guarantees finding an optimalsolution. It is relatively easy to code, even for complexproblems SA can deal with nonlinear models, unordereddata with many constraints. Its main advantages over other local searchmethods are its flexibility and its ability toapproach global optimality. It is versatile because it does not depend on anyrestrictive properties of the model. |
|
|
SA limitations: |
SA is not that much useful when the energy landscape is smooth, or there are few local minima SA is a meta-heuristic approach, so it needs a lot of choices to turn it into an actual algorithm. There is a trade-off between the quality of the solutions and the time needed to compute them. More customization work needed for varieties ofconstraints and have to fine-tune the parameters of the algorithm. The precision of the numbers used in implementation have a major effect on thequality of the result. |
|
|
ACO advantages: |
It has advantage of distributed computing. It is robust and also easy to accommodate withother algorithms. ACO algorithms have advantage over simulatedannealing and Genetic Algorithms approaches ofsimilar problems (such as TSP) when the graphmay change dynamically, the ant colonyalgorithms can be run continuously and adapt tochanges in real time. |
|
|
ACO limitations: |
Though ant colony algorithms can solve someoptimization problems successfully, we cannotprove its convergence. It is prone to falling in the local optimal solution.because the ACO updates the pheromoneaccording to the current best path. |
|
|
ACO definition: |
This approach uses the population ofChromosomes as the starting point then eachchromosome is tested against fitness using anappropriate fitness function. Then the bestchromosomes are selected and they undergo processof crossover and mutation to create new set ofchromosome. |
