This directory presents concise examples of using higher
to re-implement techniques that use unrolled gradients.
./maml-omniglot.py does few-shot Omniglot classification with MAML. For more details see the original MAML paper. Our MAML++ fork and experiments are available here.
./deep-energy-mnist.py uses SPENs/ICNNS for for MNIST classification. For more details see The original SPEN paper, End-to-End Learning for SPENs, and Input Convex Neural Networks.