Deep Learning Intermediate
AdaGrad
A sophisticated gradient descent algorithm that rescales the gradients of each parameter, effectively giving each parameter an independent learning rate.
Plain English Explanation
A sophisticated gradient descent algorithm that rescales the gradients of each parameter, effectively giving each parameter an independent learning rate. For a full explanation, see Adaptive Subgradient Methods for Online Learning and Stochastic Optimization.
How is it used?
Practitioners refer to adagrad when building, training, or evaluating machine learning systems. It appears in research papers, product documentation, and technical discussions about AI capabilities and limitations.