Glossary Library

Technical GlossaryDeep Learning

Scheduled Sampling

A method that gradually reduces teacher forcing to bring training conditions closer to inference conditions.

Scheduled sampling is a training strategy developed to mitigate exposure bias. Over time, the model is increasingly exposed to its own generated outputs rather than always seeing the correct previous output. This makes training better resemble real inference conditions. However, the speed of that transition and its effect on stability must be designed carefully.

You Might Also Like

Explore these concepts to continue your artificial intelligence journey.

Glossary Cover

Perceptron

The most basic artificial neuron model that learns a linear decision boundary through weighted inputs.

Glossary Cover

Multilayer Perceptron

A fully connected neural network structure containing multiple hidden layers.

Glossary Cover

Feedforward Neural Network

The classical family of neural networks in which information flows one-way from input to output.