Generalization transforms individual examples into a broader model that applies to unseen cases. In learning systems, it distinguishes genuine learning from memorizing previously observed examples. section 2.4
Use the distinction between memorization, induction, and deduction to assess reasoning behavior:
- A memorizing spam filter matches new messages only against stored examples, so it cannot reliably classify messages that differ from them. section 2.4
- Inductive reasoning infers broad conclusions from specific observations, but those conclusions can be false even when the observations are true. section 2.4
- Deductive reasoning applies a general hypothesis to reach a specific conclusion, assuming its premises are correct. section 2.4
Email spam classification
A learning-based spam filter should label unseen emails using patterns derived from labeled messages, rather than rerouting only exact matches found in a knowledge base. section 2.4
