Two-stage architectural fine-tuning with neural architecture search using early-stopping in image classification
This work addresses efficiency improvements in NAS for image classification, making it more practical for industry applications with limited data, though it is incremental as it builds on existing transfer learning and NAS concepts.
The paper tackles the challenge of high computational costs in neural architecture search (NAS) for image classification by proposing a two-stage architectural fine-tuning method with mutation and early-stopping, resulting in a 32.4% reduction in computational costs and a 22.3% reduction in searching costs.
In many deep neural network (DNN) applications, the difficulty of gathering high-quality data in the industry field hinders the practical use of DNN. Thus, the concept of transfer learning has emerged, which leverages the pretrained knowledge of DNNs trained on large-scale datasets. Therefore, this paper suggests two-stage architectural fine-tuning, inspired by neural architecture search (NAS). One of main ideas is mutation, which reduces the search cost using given architectural information. Moreover, early-stopping is considered which cuts NAS costs by terminating the search process in advance. Experimental results verify our proposed method reduces 32.4% computational and 22.3% searching costs.