Self-Modifying Optimizer
Tim and Tom discuss the feasibility of a self-modifying optimizer in meta learning, emphasizing the importance of gradient descent methods in optimizing meta parameters for improving learning. The conversation delves into the theoretical aspects of online meta learning and the practical implementation challenges of modifying hyperparameters directly.In this clip
From this podcast
Machine Learning Street Talk (MLST)
#49 - Meta-Gradients in RL - Dr. Tom Zahavy (DeepMind)
Related Questions