Page 471 predict the incoming flow to prevent an overflow in the tank. A fuzzy system with two input variables, the level of the sewage in the anaerobic tank and the difference in the level for two time intervals, and one output variable, the penstock aperture, is used in this task (Bailey et al. 1995). For the subtask of prediction a neural network, a fuzzy system or a fuzzy neural network can be used as shown in chapter 7.
Part B: Practical Tasks and
5. Explain the following concepts
a. Comprehensive AI
b. Multimodular hybrid system
c. Hierarchical hybrid system
d. Recurrent hybrid system
e. Hybrid connectionist production system
f. Hybrid connectionist logic programming system
g. Hybrid connectionist fuzzy production system
h. ("Pure") connectionist production system (CPS)
i. Refractoriness in a CPS
j. Random selection of a rule in a CPS
k. Selection strategy in a CPS
l. Production execution cycle in a CPS
m. Hidden Markov model
n. Dynamic time-warping method
o. Time-alignment problem in speech recognition
p. Speech segments classification into phonemic classes
6. In order to change the criterion for decision-making used at the higher level in the production rules for loan approval (see figure 6.24) to be: approve, if x10 > 0.8, or (x9 > 0.9 and x10 > 0.6), or (x8 > 0.8 and x9 > 0.8 and x10 > 0.7), do you need to retrain the neural network?
10. List the advantages and disadvantages of using hybrid connectionist systems for character recognition.
11. List all the steps to be taken for developing a hybrid, phoneme-based speech recognition system as part of an intelligent human computer interface (see figures 1.20 and 6.14) to a database containing clinical data.
2. For the task chosen from the list below, create a multimodular block diagram of a possible solution to the problem.
3. Choose appropriate techniques for solving each subproblem represented as a module. What alternatives are there for each of them?