# Loss Function

> Source: https://sukruyusufkaya.com/en/glossary/kayip-fonksiyonu
> Updated: 2026-05-13T21:10:43.350Z
> Type: glossary
> Category: yapay-zeka-temelleri
**TLDR:** A mathematical function that quantifies the difference between model predictions and true values and guides the training process.

<p>A loss function is the fundamental mechanism that numerically expresses how wrong a model is. During training, the optimization process tries to reduce this function, meaning it attempts to shrink the gap between predictions and true outcomes. But the key point is that a loss function is not merely a technical error measure. It also defines what the system treats as an “important mistake.” For that reason, choosing the right loss function can be as important as choosing the right model. A model trained against the wrong objective may look technically optimized while still producing outcomes misaligned with business goals.</p>