6 terms
Showing all terms starting with W
A numerical parameter in a neural network adjusted during training to minimise prediction error.
Using AI to automatically execute sequences of tasks across applications triggered by rules or events without manual intervention.
A training schedule that starts with a very low learning rate and gradually increases it, preventing large early updates from destabilising training.
Automated extraction of data from websites using AI to parse, clean, and structure the collected information for training or analysis.
A regularisation technique that penalises large model weights during training to prevent overfitting and improve generalisation.
An internal representation an AI agent builds of its environment, enabling it to simulate future states and plan actions without direct interaction.
