Momentum Deep Learning
noun
Definition: A gradient-based optimization technique that updates parameters using both the current gradient and an exponentially weighted contribution [Bakanach 2023/2024].
Example in context: “The momentum iterative method (MIM) is an extension of the BIM attack that introduces momentum to accelerate the convergence toward adversarial examples.” [Cali et al. 2024]
Related terms: gradient descent, SGD with momentum, optimizer