AI · Theory

Forward and Backward Chaining

Theory

40/215

Forward and Backward Chaining

Forward chaining is data-driven: new facts fire rules, more facts appear, until the query is known. Backward chaining is goal-driven: start from what you want, ask what would prove it, recurse. Expert systems use both.

Medical: lots of symptoms arriving → forward. “Does the patient have measles?” → backward. Horn clauses keep both tractable.

Forward and Backward Chaining — output — forward derived {'dirt', 'suck'}. Backward starts from the goal suck.

Exam tip

One situation for each chaining style.

Example

# Forward vs backward
facts = {"dirt"}
rules = [("dirt", "suck")]
# forward: facts fire rules
derived = set(facts)
for cond, act in rules:
    if cond in derived:
        derived.add(act)
print("forward derived", derived)
# backward: start from goal suck
print("backward asks: what proves suck? → dirt, which is known")

Forward and Backward Chaining — output: forward derived {'dirt', 'suck'}. Backward starts from the goal suck.

Short notes

DefForward : data-driven.
RuleBackward : goal-driven.

# Forward vs backward facts = {"dirt"} rules = [("dirt", "suck")] # forward: facts fire rules derived = set(facts) for cond, act in rules: if cond in derived: derived.add(act) print("forward derived", derived) # backward: start from goal suck print("backward asks: what proves suck? → dirt, which is known")

Forward and Backward Chaining

Example

Short notes

Questions

Forward and Backward Chaining

Example

Short notes

Questions