Effect of Optimizer, Initializer, and Architecture of Hypernetworks on Continual Learning from Demonstration
This is an incremental study for robotics researchers working on continual learning from human demonstrations.
The paper investigates how different optimizers, initializers, and network architectures affect hypernetworks in continual learning from demonstration for robots, finding that adaptive optimizers work well but specialized initializers do not improve performance, and stable hypernetworks are robust to architectural changes.
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. In this paper, we perform an exploratory study of the effects of different optimizers, initializers, and network architectures on the continual learning performance of hypernetworks for CLfD. Our results show that adaptive learning rate optimizers work well, but initializers specially designed for hypernetworks offer no advantages for CLfD. We also show that hypernetworks that are capable of stable trajectory predictions are robust to different network architectures. Our open-source code is available at https://github.com/sebastianbergner/ExploringCLFD.