Effect of Optimizer, Initializer, and Architecture of Hypernetworks on Continual Learning from Demonstration

In continual learning from demonstration (CLfD), a robot learns a sequence of real-world motion skills continually from human demonstrations. Recently, hypernetworks have been successful in solving this problem.