olemus
ISO/IEC 22989:
andmed, mida kasutatakse kandidaatmudelite sooritusvõime võrdlemiseks;
(i) valideerimisandmed ei ole testandmed ja üldiselt ka mitte treenimisandmed; kui aga andmeid ei piisa kolmeks otstarbeks (treening, valideerimine, test) eraldi, jagatakse andmed ainult kaheks -- testimisandmestikuks ning treeningu ja valideerimise andmestikuks; sellisel juhul kasutatakse kaheotstarbelisest anmestikust eraldi treeningu- javalideerimisandmestike genereerimiseks tavaliselt ristvalideerimist või statistilist teisendust
(ii) valideerimisandmeid saab kasutada hüperparameetrite häälestamiseks või mingi algoritmilise valiku kehtestuseks kuni ekspertsüsteemile teatud reegli lisamiseni
= data used to compare the performance of different candidate models
Note 1. Validation data is disjoint from test data and generally also from training data. However, in cases where there is insufficient data for a three-way training, validation and test set split, the data is divided into only two sets – a test set and a training or validation set. Cross-validation or bootstrapping are common methods for then generating separate training and validation sets from the training or validation set.
Note 2. Validation data can be used to tune hyperparameters or to validate some algorithmic choices, up to the effect of including a given rule in an expert system.

ülevaateid
https://www.techtarget.com/whatis/definition/validation-set
https://en.wikipedia.org/wiki/Training,_validation,_and_test_data_sets
https://www.larksuite.com/en_us/topics/ai-glossary/validation-data