论文标题
RPR：训练的随机分区放松；二进制和三元体重神经网络
RPR: Random Partition Relaxation for Training; Binary and Ternary Weight Neural Networks
论文作者
论文摘要
我们提出随机分区弛豫（RPR），这是一种对神经网络重量对二进制（+1/-1）和三元（+1/0/-1）值进行强量化的方法。从预先训练的模型开始，我们将权重量化，然后将它们的随机分区放松到其连续的值中，以便在重新量化它们并切换到另一个重量分区以进行进一步适应之前进行重新训练。我们使用基于SGD的培训方法可以轻松地集成到现有框架中，展示了Googlenet和Resnet-50的最先进的二进制和三元重量网络，以及RESNET-18和RESNET-50的竞争性能。
We present Random Partition Relaxation (RPR), a method for strong quantization of neural networks weight to binary (+1/-1) and ternary (+1/0/-1) values. Starting from a pre-trained model, we quantize the weights and then relax random partitions of them to their continuous values for retraining before re-quantizing them and switching to another weight partition for further adaptation. We demonstrate binary and ternary-weight networks with accuracies beyond the state-of-the-art for GoogLeNet and competitive performance for ResNet-18 and ResNet-50 using an SGD-based training method that can easily be integrated into existing frameworks.
