Appendix for Evaluating Efficient Performance Estimators of Neural Architectures