86. c. Just as replication complicates concurrency control, it can affect scalability. The major concern in scalability is determining the effect of increased scale on client performance. Additional storage sites increase the amount of work servers must do to maintain a consistent state of the file system. Similarly, clients in a replicated file system may have more work to do when they make file updates. For this reason, both clients and servers share portions of system management work.
Fault-tolerant mechanisms, availability, and recoverability are incorrect. Replicated servers have a positive impact on system availability and recoverability. If the primary server fails, the replicated server takes over, thus making the system available to system users. Recovery protocols help both servers and clients recover from system failures. Fault-tolerant mechanisms such as disk mirroring and disk duplexing help in recovering from a system failure. They all have a positive effect.
87. Which of the following statements about expert systems is not true?
a. Expert systems are aimed at solving problems using an algorithmic approach.
b. Expert systems are aimed at solving problems that are characterized by irregular structure.
c. Expert systems are aimed at solving problems characterized by incomplete information.
d. Expert systems are aimed at solving problems characterized by considerable complexity.
87. a. Expert systems are aimed at problems that cannot always be solved using a purely algorithmic approach. These problems are often characterized by irregular structure, incomplete or uncertain information, and considerable complexity.
88. In the context of expert systems, a heuristic is not a:
a. Rule of thumb
b. Known fact
c. Known procedure
d. Guaranteed procedure
88. d. A heuristic is a rule of thumb, a known fact, or even a known procedure that can be used to solve some problems, but it is not guaranteed to do so. It may fail. Heuristics can be conveniently regarded as simplifications of comprehensive formal descriptions of real-world systems. These heuristics are acquired through learning and experience.
89. The architecture of an expert system does not include which one of the following?
a. Knowledge base
b. Computing environment
c. Inference engine
d. End user interface
89. b. The computing environment consists of hardware, programming languages, editors and compilers, file management facilities, browsing program code, debugging and tracing program execution, and graphic programming. This computing environment is outside the expert systems architecture because it can change from one organization to another.
On the other hand, knowledge base, inference engine, and end user interface are integral parts of expert systems architecture. Knowledge is stored in the knowledge base using symbols and data structures to stand for important concepts. The symbols and data structures are said to represent knowledge. A software module called the inference engine executes inference procedures. If the user of the expert system is a person, communications with the end user are handled via an end user interface.
90. Expert systems differ from conventional systems in all the following except:
a. Expert system knowledge is represented declaratively.
b. Expert system computations are performed through symbolic reasoning.
c. Expert system knowledge is combined into program control.
d. Expert systems can explain their own actions.
90. c. Expert system programs differ from conventional systems in four important ways. First, knowledge is separated from program control; the knowledge base and inference engine are separate. Second, knowledge is represented declaratively. Third, expert systems perform computation through symbolic reasoning. And finally, expert systems can explain their own actions.
91. Which of the following categories of problem-solving activity is best suited to expert systems?
a. Tasks based on a limited domain
b. Tasks based on common sense knowledge
c. Tasks requiring perceptual knowledge
d. Tasks based on creativity
91. a. The size of completed expert systems is often large, consisting of hundreds or thousands of rules. If the task is too broad, the development effort may take an inordinate amount of time, or even be impossible. Two important guidelines on evaluating the scope and size of the problem include the task must be narrowly focused and the task should be decomposable. In other words, expert system tasks should be based on a limited domain.
