首页|Efficient neural network using pointwise convolution kernels with linear phase constraint
Efficient neural network using pointwise convolution kernels with linear phase constraint
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
Elsevier
In current efficient convolutional neural networks, 1 × 1 convolution is widely used. However, theamount of computation and the number of parameters of 1 × 1 convolution layers account for a large partof these neural network models. In this paper, we propose to use linear-phase pointwise convolution kernels(LPPC kernels) to reduce the computational complexities and storage costs of these neural networks.We design four types of LPPC kernels based on the parity of the number of input channels and symmetryof the weights of the pointwise convolution kernel. Experimental results show that Type-Ⅰ LPPC kernelscan compress some popular networks better with a small reduction in accuracy than the other types ofLPPC kernels. The LPPC kernels can be used as new 1 × 1 convolution kernels to design efficient neuralnetwork architectures in the future. Moreover, the LPPC kernels are friendly to low-power hardwareaccelerator design to achieve lower memory access cost and smaller model size.
NETL
NSTL
Elsevier
