järeldusotsing

Järeldusotsing, also known as backward chaining, is a goal-driven inference method used in artificial intelligence and logic programming. It begins with a potential conclusion or goal and works backward to find evidence or facts that support it. In essence, it asks "Can this goal be true?" and then tries to find a way to prove it.

This process involves breaking down the desired conclusion into subgoals. Each subgoal is then treated as a new goal, and the system attempts to find rules or facts that can satisfy it. If a subgoal can be satisfied by a known fact, that branch of the search is complete. If it can be satisfied by another rule, that rule becomes a new subgoal, and the process continues recursively.

Järeldusotsing is commonly employed in expert systems and theorem provers. For example, if a system's goal is to diagnose a patient's illness, it might start with the possible diagnosis and then look for symptoms and test results that are consistent with it. This contrasts with forward chaining, which starts with available data and works forward to derive conclusions. The choice between järeldusotsing and forward chaining often depends on the nature of the problem and the direction in which the inference is most efficiently performed.