A feature is a property, attribute, or characteristic of an entity that has been selected, engineered, or extracted from raw data to serve as a predictive input. Features are the units of representation that models operate on: they translate the messy structure of the world into numbers, categories, or tokens that a learning algorithm can process. Feature engineering is the practice of designing and transforming features; feature selection is the practice of choosing the subset that matters. Features are not natural givens — they are the result of human decisions about what to measure, how to encode it, and what to discard. Parameters and persistence: features are defined by their domain semantics, their data type (numeric, categorical, text, image patch), their scale, and their relationship to the target variable; they persist through versioned feature stores, schema registries, and documentation that track how each feature was computed, from what source, and with what transformation pipeline. [formal: proprietas | substrate: mind | horizon: a project | explicit: yes | epoch: 0.01]
Accepted ontology entry
feature
A feature is a property, attribute, or characteristic of an entity that has been selected, engineered, or extracted from raw data to serve as a predictive input. Features are the units of representation that models operate on: they transla…
Definition
Why it is in scope
A feature is a human-made property or characteristic extracted or selected from raw data to serve as input for a model or analysis. It is the bridge between raw observations and machine-readable structure, designed to carry predictive signal.
Names and aliases
- featureen · CANONICAL
Relations from this entry
- cmrc361dn0082a9g46bcbfpitINSTANCE_OF →
TESTED INSTANCE_OF: A feature IS a specific kind of property — a distinguishing attribute or characteristic of a thing. Specific-to-general tested. 'Property' as the nearest kind.
Relations to this entry
- cmsfk0hj307eb3vv3cf44f9sy← DEPENDS_ON
TESTED: remove features and a decision boundary stops operating — it IS a surface in feature space. Without features, there is no space, no boundary, no classification function. This is not merely sayability; the concept's operation collapses.
- cmspmjaky05vwjlss1p9r2wo0← INSTANCE_OF
A cepstral coefficient is a specific type of signal feature — a numerical representation extracted from a signal that captures distinctive characteristics. A competent speaker would call a cepstral coefficient 'a feature.' INSTANCE_OF direction (specific→general) is correct.
- cmsr8ozye00oekp53923ksacd← INSTANCE_OF
Delta cepstral coefficients are a specific kind of feature: they are extracted acoustic descriptors that compact vocal tract information into a feature vector for downstream classification. A competent speaker in signal processing would call them 'acoustic features.' Per Law 9: specific→general INSTANCE_OF.
- cmsqwdhy2004jax3hdwf3yowy← INSTANCE_OF
Mel-spectrogram is a specific kind of feature: it represents signal energy distribution across mel-frequency bins and time frames, forming a compact acoustic descriptor used directly in classification and recognition pipelines. A competent speaker would call it 'an acoustic feature' or 'a feature representation.' Per Law 9: specific→general INSTANCE_OF.
- cmsreo7w50175kp53txs4kt3o← INSTANCE_OF
Chroma-feature IS a specific kind of feature: a 12-dimensional audio representation vector. A competent speaker would say 'chroma feature is a feature.' Files against nearest kind — no rung skipped.
- cmss0y7r700vsh7yuppfym4fz← INSTANCE_OF
Audio-feature IS a specific kind of feature — a quantifiable property from audio signals for analysis. A competent speaker calls it 'a type of feature'. Files against the nearest kind per Law 9.
Record identity
- Created
- Aug 4, 2026, 5:42 PM UTC
- Content hash
- 87817d28adfff37b0fa33e47d95a54de2d8e1c164b301a20aa548f36724dcd8b