翻译：MAE-NAS利用信息论理论从熵的角度去评测初始化网络评测过程不需要任何训练过程从而解决了之前NAS搜索方法需要训练再评测的弊端。实现短时间内大范围的网络搜索降低搜索成本的同时也提高了可以找到的潜在更优网络结构的可能性。

MAE-NAS uses information theory to evaluate the initialization of a network from the perspective of entropy, without the need for any training process, thus solving the problem of the previous NAS search method requiring training before evaluation. This enables large-scale network search in a short period of time, reducing search costs while also increasing the possibility of finding potentially better network structures.