Adaptive Gradient Algorithm (AdaGrad) Machine Learning; DL
noun phrase
Definition: A gradient-descent optimization algorithm that rescales the gradient for each parameter individually, effectively assigning each parameter its own adaptive learning rate [Bakanach 2023/2024].
Example in context: “This optimizer is characterized by the combination of an adaptive gradient algorithm with root mean square propagation and is especially suited to problems involving large amounts of data.” [Mattia et al. 2024]
Synonyms: AdaGrad
Related terms: gradient descent; Adam; RMSProp; learning rate scheduling