A hyperparameter is a human-made concept in machine learning describing configuration variables whose numerical or categorical values are set by a practitioner before the training process begins, governing the model architecture, optimization dynamics, or regularization strength. Examples include learning rate, number of hidden layers, tree depth, dropout rate, and batch size. Unlike model parameters (weights and biases) that are learned automatically from data during training, hyperparameters are fixed during training and must be selected through manual specification, search procedures (grid search, random search, Bayesian optimization), or domain expertise. Hyperparameters persist through configuration files, experiment tracking systems, version control of scripts, and scientific publication of methodology. [formal: hyperparametrum | substrate: mind | horizon: hours | explicit: yes | epoch: 1.00]
Accepted ontology entry
hyperparameter
A hyperparameter is a human-made concept in machine learning describing configuration variables whose numerical or categorical values are set by a practitioner before the training process begins, governing the model architecture, optimizat…
Definition
Why it is in scope
A human-made concept in machine learning describing configuration variables whose values are set by a practitioner before the training process begins — as opposed to parameters learned during training. Examples include learning rate, number of hidden layers, regularization strength, and batch size. The concept exists as a design choice space that practitioners navigate to optimize model performance.
Names and aliases
- hyperparameteren · CANONICAL
Relations from this entry
- cmru5nrqe003sr671qxjyxhiqDEPENDS_ON →
Removal test: remove model — do hyperparameters stop operating? Yes. Hyperparameters are configuration values set before training to control how a model learns. Without models, hyperparameters have no operation or meaning. Constitutive dependency.
- cmrvsdf2u027u2ceisomaun7qINSTANCE_OF →
TESTED INSTANCE_OF: A hyperparameter IS a specific kind of variable — a configurable parameter whose value is set before training begins. Variable is the nearest general kind (not jumping to 'parameter' which doesn't exist yet).
- cmrg0scos00ef2a1nklfvbk7xDERIVED_FROM →
Hyperparameters are configuration values whose design and purpose derive from machine learning practice. Machine learning existed first and created the concept of tuning parameters beyond the core algorithm — a hyperparameter is a concept that emerged from the practice of training ML models.
Relations to this entry
- cmslplzei07tonobp3lnlzpuz← INSTANCE_OF
Batch size is a specific kind of hyperparameter in machine learning. A competent speaker calls batch size 'a hyperparameter.' It is set before training to control the number of samples per gradient update — fitting the general definition of a hyperparameter (a configuration set before the learning process begins).
- cmsn65s8b034k1q13zxnf04p5← INSTANCE_OF
Temperature is a specific kind of hyperparameter — a scalar configurable parameter that controls model behavior during inference. A competent speaker would call temperature 'a hyperparameter' per Law 9.
Record identity
- Created
- Aug 4, 2026, 4:13 PM UTC
- Content hash
- 842ce15823f1f99a401cecc476732fca5d144e20328724acf4eb17e8eb4ceee1