Routing is a key stage in current IC design. Due to its hardness in optimization, several machine learning algorithms have been developed recently. But they often suffer from the issues like high time complexity for training and large demand of training data. Moreover, modern routing usually needs to consider various optimization objectives (e.g., the total wirelength or the maximum time delay). It is prohibitively expensive if we always train a new model from scratch to adapt each encountered new routing objective. In this paper, we introduce a novel Coreset-based Transfer Learning (CoTL) framework that can significantly reduce the training time and the amount of newly added training data. The key part of our framework relies on a novel sampling idea called "model-guided coreset", which can yield 5-8X reduction on the training time with preserving comparable routing quality to the state-of-the-art learning based approaches.

In recent years, there has been an explosion of research into developing robust deep neural networks against adversarial examples. As one of the most successful methods, Adversarial Training (AT) has been widely studied before, but there is still a gap to achieve promising clean and robust accuracy for many practical tasks. In this paper, we consider the AT problem from a new perspective which connects it to catastrophic forgetting in continual learning (CL). Catastrophic forgetting is a phenomenon in which neural networks forget old knowledge upon learning a new task. Although AT and CL are two different problems, we show that they actually share several key properties in their training processes. Specifically, we conduct an empirical study and find that this forgetting phenomenon indeed occurs in adversarial robust training across multiple datasets (SVHN, CIFAR-10, CIFAR-100, and TinyImageNet) and perturbation models ($\ell_{\infty}$ and $\ell_{2}$). Based on this observation, we propose a novel method called Adaptive Multi-teachers Self-distillation (AMS), which leverages a carefully designed adaptive regularizer to mitigate the forgetting by aligning model outputs between new and old ``stages''. Moreover, our approach can be used as a unified method to enhance multiple different AT algorithms. Our experiments demonstrate that our method can significantly enhance robust accuracy and meanwhile preserve high clean accuracy, under several popular adversarial attacks (e.g., PGD, CW, and Auto Attacks). As another benefit of our method, we discover that it can largely alleviate the robust overfitting issue of AT in our experiments.