Glossary term
Glossary term
Foundations
A system that always returns the same output for a given input. For example, the ReLU function is deterministic because:
When the input is negative, the output is always 0.
When the input is nonnegative, the output always equals the input.
By contrast, a function that returns a random number each time it is called is nondeterministic.
Deterministic systems are generally much easier to test than nondeterministic systems.
LLMs are usually nondeterministic; that is, the LLM's response to the same prompt often differs.
Created for this library
A bank's model risk team requires deterministic inference for its production credit scorecard so the same input always produces the same decision.
An ML platform team enforces deterministic data preprocessing so a model produced by retraining on the same input bytes is byte-identical to the previous run.
A research lab freezes random seeds in evaluation scripts so the deterministic results can be reproduced in the published paper.
Definition source: Google for Developers Machine Learning Glossary | Creative Commons Attribution 4.0 License