Glosarium AI

📖

istilah

Pre-trained Model

A neural network that has already been trained on a large dataset (like ImageNet) for a generic task, whose weights and architecture are then reused for more specific tasks.

📖

istilah

Frozen Layers

Technique consisting of disabling the weight (gradient) updates of certain layers of a pre-trained model during fine-tuning, in order to preserve the learned representations and prevent overfitting.

📖

istilah

Differential Learning Rate

Fine-tuning method applying distinct learning rates to the different layers of the model, often higher for the upper (specific) layers and lower for the lower (generic) layers.

📖

istilah

Classifier Head

The final layers of a pre-trained model, typically one or more fully connected layers, which are replaced or re-trained to adapt to the specific classes of the new task.