Você está na página 1de 4

TOPIC 10 INFERENCE

W 143

Topic

Inference

10
LEARNING OUTCOMES
By the end of this topic, you should be able to: 1. Explain the de nition of inference; 2. State modus ponens with further details; 3. Identify ways to achieve a resolution; and 4. Find out how a non-resolution is achieved.

X INTRODUCTION
You have already learnt about the types of reasoning used by humans, which are Deductive Reasoning, Inductive Reasoning, Abductive Reasoning, Analogical Reasoning, Common Sense Logical Reasoning and Non-monotonic Reasoning. Other types of reasoning that have also been taught are Model-based Reasoning and Case-based Reasoning. These types of reasoning are very important in forming rules in knowledge base so that inference process can be carried out in a more organised and ef cient way. Intelligent systems will, indirectly, model human reasoning using inference techniques.

144

TOPIC 10 INFERENCE

10.1

WHAT IS INFERENCE?

Figure 10.1: Intelligent system process

Inference can be said as: A process used by an intelligent system to produce or infer new information from existing information. Before this, Topic 7 had explained about inference used in Predicate Logic Knowledge Representation. Topic 7 had also provided details on modus ponens and resolutions. In this topic, inference for intelligent system applications will be explained again.

ACTIVITY 10.1
Can you differentiate between the roles of reasoning and inference?

10.2

INFERENCE CONCEPT

Among the features of inference engine that must be considered are: (a) What are the questions to know? (b) How to search a Knowledge Base?

TOPIC 10 INFERENCE

W 145

(c) How to select and re a rule from other rules? (d) How is the result in uenced by the search?

10.3

UNCERTAINTY INFERENCE TECHNIQUE

Another inference technique that can be used is Uncertainty Inference. An expert will provide inaccurate knowledge in the form of rules by stating a value estimate. Inaccurate knowledge for a basic set of events can be directly used for inferring small cases. When running an inference engine, an expert can change the input after seeing the results. Normally, situations are related to one another. Hence, there is a need for global values for a system. Main integration methods that can be used are Bayesian Probability, Proving Theory, Certainty Factor and Fuzzy Set. Uncertainty problem is a serious problem. However, uncertainty can be represented in numeric, graphic or symbolic. For example, numeric uncertainty representations are: (a) Scale (0-1, 0-100) (b) 0 = total uncertainty (c) 1 or 100 = total certainty Another problem that needs to be resolved is cognitive biases. Humans can be inconsistent at different times. Therefore, uncertainty inference is important in getting exible and consistent answers.

146

TOPIC 10 INFERENCE

ACTIVITY 10.2
(a) (b) (c) (d) What is meant by inference? What is the purpose of carrying out resolution? give an example of uncertainty inference. What is meant by resolution theorem? Why is negation of statement needed? (e) Assume the following is an axiom set at state-0. List other state when using Modus Ponens. (i) E o ( (ii) E o ( (iii) E o ( (vi) E o ( (v) E o ( (vi) E o ( (vii) E

SUMMARY
z

In summary, knowledge must be processed or reasoned and a computer program can access knowledge for inference. Inference engine is a controlled program that contains rules interpreter and directs search in order to nd answers in the knowledge base.

Você também pode gostar