A.3 Analysis on the Ratio of Discarded Outliers As we discussed in the subsection of outliers clamp for activations, the value Table 3: Analysis on the ratio of discarded outliers in activations. Besides, the comparisons with more CNN quantization methods are also supplemented. In Figure 1, we visualize the histogram of the weights and activations in AdderNet. Our AdderNet quantization method has one major limitation: as the number of bits decreases, the accuracy loss of the quantization model will increase. As for the societal impacts, the proposed quantization method can further reduce the energy consumption of AdderNet with a lower quantized accuracy loss.

A.3 Analysis on the Ratio of Discarded Outliers As we discussed in the subsection of outliers clamp for activations, the value Table 3: Analysis on the ratio of discarded outliers in activations. Besides, the comparisons with more CNN quantization methods are also supplemented. In Figure 1, we visualize the histogram of the weights and activations in AdderNet. Our AdderNet quantization method has one major limitation: as the number of bits decreases, the accuracy loss of the quantization model will increase. As for the societal impacts, the proposed quantization method can further reduce the energy consumption of AdderNet with a lower quantized accuracy loss.

Adder Neural Network (AdderNet) provides a new way for developing energy-efficient neural networks by replacing the expensive multiplications in convolution with cheaper additions (i.e., L1-norm). To achieve higher hardware efficiency, it is necessary to further study the low-bit quantization of AdderNet. Due to the limitation that the commutative law in multiplication does not hold in L1-norm, the well-established quantization methods on convolutional networks cannot be applied on AdderNets. Thus, the existing AdderNet quantization techniques propose to use only one shared scale to quantize both the weights and activations simultaneously. Admittedly, such an approach can keep the commutative law in the L1-norm quantization process, while the accuracy drop after low-bit quantization cannot be ignored.