▸创新点1：基于特征积的卷积方法（FIBC）是一种算法创新，通过将卷积操作转换为特征积分运算，显著降低了算术复杂度，提升了计算效率，适用于二进制和三元权重神经网络。
▸创新点2：核变换特征重构卷积方法（KTFR）是一种算法创新，通过核变换和特征重构技术，有效减少了卷积操作中的冗余计算，进一步优化了计算性能。
▸创新点3：分层负载均衡机制（HLBM）是一种系统创新，通过动态分配计算资源并消除零值计算，提高了资源利用率和能效，实测能效提升1.25倍。
▸创新点4：联合优化方法是一种系统创新，通过为每一层卷积搜索最优计算模式，实现了整体性能的最大化，能效提升2.24倍。
Abstract
Due to less memory requirement, low computation overhead and negligible accuracy degradation, deep neural net- works with binary/ternary weights (BTNNs) have been widely employed on low-power mobile and Internet of Things (IoT) devices with limited storage capacity. Some hardware imple- mentations have been proposed to accelerate the inference of BTNNs by utilizing the multiplication-free feature. However, some implicit characteristics in BTNN convolution, such as high arithmetic complexity and numerous redundant opera- tions, are never considered. In this paper, we propose four optimization techniques to fully exploit these features. First, a feature-integral-based convo lution (FIBC) method is proposed to reduce the arithmetic complexity of convolutional layers. Second, a kernel-transformation -feature-reconstruction (KTFR) convolution method is presented to remove redundant operations in BTNN convolution. Third, a hierarchical load-balancing mech- anism (HLBM) is designed to eliminate zero value computation and improve resource utilization. Finally, a joint optimization approach for convolutional layers is proposed to search opti- mal calculation pattern for each layer. Based on the pro- posed four techniques, we design a reconfigurable processor in a 28-nm CMOS technology to accelerate the inferences of BTNNs. The four proposed techniques improve energy efficiency by 2.07 ×,1 . 6 5×,1 . 2 5×, and 2.24 × for BTNNs respectively, compared with the baseline implementation which