Equal Bits: Enforcing Equally Distributed Binary Network Weights
This work addresses a specific bottleneck in binary neural networks by enabling precise control over weight distributions, offering incremental improvements in efficiency and optimization for resource-constrained AI applications.
The paper tackled the problem of precisely controlling the distribution of binary network weights during training to achieve maximum entropy, which prior methods could not guarantee. They demonstrated that using optimal transport for quantization can enforce any bit ratio, including equal ratios, leading to optimization benefits and competitive performance compared to state-of-the-art binarization methods.
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.