GAE for Generalized Advantage Estimation is an actor critic model that uses some tricks to improve training of the model.
GAE uses lambda returns a generalization of n-steps where the model does’nt use only the n-th step but a combination of each n-th first steps. The parameter \(\lambda\) defines the weights associated with each step.
\(\lambda=0\) is equivalent to TD estimate and \(\lambda=1\) is equivalent to MC estimate. Every \(\lambda\) in between is a mix.
See: