Equal Bits: Enforcing Equally Distributed Binary Network Weights
DOI:
https://doi.org/10.1609/aaai.v36i2.20039Keywords:
Computer Vision (CV)Abstract
Binary networks are extremely efficient as they use only two symbols to define the network: {+1, −1}. One can make the prior distribution of these symbols a design choice. The recent IR-Net of Qin et al. argues that imposing a Bernoulli distribution with equal priors (equal bit ratios) over the binary weights leads to maximum entropy and thus minimizes information loss. However, prior work cannot precisely control the binary weight distribution during training, and therefore cannot guarantee maximum entropy. Here, we show that quantizing using optimal transport can guarantee any bit ratio, including equal ratios. We investigate experimentally that equal bit ratios are indeed preferable and show that our method leads to optimization benefits. We show that our quantization method is effective when compared to state-of-the-art binarization methods, even when using binary weight pruning. Our code is available at https://github.com/liyunqianggyn/Equal-Bits-BNN.Downloads
Published
2022-06-28
How to Cite
Li, Y., Pintea, S.-L., & Gemert, J. C. van. (2022). Equal Bits: Enforcing Equally Distributed Binary Network Weights. Proceedings of the AAAI Conference on Artificial Intelligence, 36(2), 1491-1499. https://doi.org/10.1609/aaai.v36i2.20039
Issue
Section
AAAI Technical Track on Computer Vision II