Hands-On Mathematics for Deep Learning

Model-agnostic meta learning

Model-Agnostic Meta Learning (MAML) is an optimization algorithm that can work on any type of neural network that is trained using gradient descent. Suppose we have a model, f, with parameters, θ, and a task, τ_i, that has a corresponding dataset . Then, we can make updates to the model using a single or several gradient descent steps. A single step in this algorithm works as follows:

The preceding step learns to optimize a single task, but we would like to optimize multiple tasks. So, we can change the task to find ...