基于增強PSO算法的神經(jīng)架構(gòu)搜索方法在工業(yè)圖像分類中的應(yīng)用研究發(fā)表在information sciences
Neural architecture search using an enhanced particle swarm optimization algorithm for industrial image classification
Highlights
- •
An enhanced PSO algorithm incorporating dynamic ring neighbourhood topology and swarm entropy mutation mechanisms is proposed.
- •
We develop a two-level binary particle encoding scheme to improve search efficiency.
- •
We use MBConv modules as basic building blocks and replace their original SE attention with CBAM mechanisms.
- •
Addressing the computational bottleneck in NAS, we propose a low-fidelity evaluation strategy.
聲明：本內(nèi)容系學者網(wǎng)用戶個人學術(shù)動態(tài)分享，不代表平臺立場。
學者網(wǎng)
評論 0