# Training Split (Edge AI Glossary)

## metadata
- term: Training Split
- slug: training-split
- letter: T
- route: /resources/glossary/training-split
- canonical: https://klyff.com/resources/glossary/training-split

## Definition & Context
A training split is the process of dividing datasets into subsets for training, validation, and testing in machine learning, ensuring Edge AI models learn accurately and perform...

## Details
## Training Split

#### What is Training Split?

A training split is the process of dividing datasets into subsets for training, validation, and testing in machine learning, ensuring Edge AI models learn accurately and perform reliably on-device. Also known as data partitioning, it optimizes model learning and evaluation.

#### Why Is It Used?

Training splits are essential to prevent overfitting, validate model performance, and ensure that AI models deployed on edge devices operate efficiently and safely in real-world conditions.

#### How Is It Used?

Data is divided into portions:

 
- **Training set:** Used to teach the AI model patterns.
- **Validation set:** Used to tune hyperparameters and optimize performance.
- **Test set:** Evaluates the final model’s accuracy on unseen data. 

#### Types of Training Split

 
- **Standard split (e.g., 70/15/15)**: Most common in Edge AI projects.
- **K-Fold cross-validation**: Rotates subsets to ensure model robustness.
- **Time-based split**: Useful for sequential or IoT streaming data. 

#### Benefits of Training Split

 
- Improves model accuracy on edge devices.
- Reduces risk of overfitting and performance bias.
- Enables efficient resource usage for on-device computation.
- Enhances reliability in real-world AI applications.