Chongjun Tu, Peng Ye, Weihao Lin, Hancheng Ye, Chong Yu...
TL;DR提出一种基于 Bi-level Data Pruning (BDP) 范式的新型神经架构搜索优化方法,通过逐步修剪不适合的样本来降低搜索成本,并实现超过 50% 的性能提升。
Abstract
Improving the efficiency of neural architecture search (NAS) is a challenging
but significant task that has received much attention. Previous works mainly
adopted the differentiable architecture search (DARTS) an