Skip to content

Concept Glossary: K

This page lists English glossary entries for this letter. Entry bodies are assembled from term source files.

k-NN, k-nearest neighbors

  • Meaning: k-NN is a model that finds the k nearest stored examples to a new input and predicts from their majority vote or average. It relies directly on similarity to nearby cases rather than learning a complex global formula.
  • Why it matters: k-NN gives a direct case-based intuition for machine learning. It shows why distance, feature scale, and the chosen number of neighbors can strongly change model behavior.
  • Related concepts: nearest neighbor, distance, scale
  • Core Section: P4-12.1
  • Appears in: P4-12.2

key

  • Meaning: A key is the name or identifier used to find a value in a mapping structure such as a dictionary. Unlike a list index, which points to position, a key points by meaningful label.
  • Why it matters: Configuration, JSON, and metadata often make sense only when the key name is read with the value. The same value can mean different things when attached to different keys.
  • Related concepts: dictionary, mapping, value
  • Core Section: P2-8.3
  • Appears in: P2-8.4, P2-9.4

knowledge acquisition

  • Meaning: Knowledge acquisition is the process of turning expert judgment and domain knowledge into rules or structures a system can use. It translates human know-how into machine-usable form.
  • Why it matters: Rule-based systems were difficult not only because of implementation, but because good knowledge had to be extracted, checked, encoded, and maintained over time.
  • Related concepts: expert system, knowledge base, knowledge representation
  • Core Section: P1-3.1

knowledge base

  • Meaning: A knowledge base is a structured store of rules, facts, relationships, and domain information that a system can consult. It separates knowledge materials from the procedure that applies them.
  • Why it matters: The concept helps readers see why rule-based systems are more than long lists of conditional code. The knowledge base stores materials, while an inference process uses them.
  • Related concepts: rule-based system, knowledge representation, inference engine
  • Core Section: P1-3.1
  • Appears in: P1-2.1, P1-3.3

knowledge representation

  • Meaning: Knowledge representation is the work of organizing facts, relationships, rules, and constraints in a form that a computer can inspect, compare, and reason over.
  • Why it matters: Symbolic AI depends on not only what knowledge exists, but how that knowledge is written. Tables, rules, graphs, and logical forms support different questions and reasoning procedures.
  • Related concepts: symbolic AI, rule-based system
  • Core Section: P1-2.1
  • Appears in: P1-2.2

KV cache

  • Meaning: A KV cache stores previously computed key and value representations for earlier tokens so they can be reused during later token generation instead of recomputed from scratch.
  • Why it matters: Long-context generation can become slow if every step recalculates the whole past context. KV cache connects generation speed and cost to how a system reuses previous computation.
  • Related concepts: query-key-value, QKV, context window, long-context
  • Core Section: P6-4.3
  • Appears in: P5-14.2