TY - JOUR T1 - Quantization and Training of Low Bit-Width Convolutional Neural Networks for Object Detection AU - Yin , Penghang AU - Zhang , Shuai AU - Qi , Yingyong AU - Xin , Jack JO - Journal of Computational Mathematics VL - 3 SP - 349 EP - 359 PY - 2018 DA - 2018/09 SN - 37 DO - http://doi.org/10.4208/jcm.1803-m2017-0301 UR - https://global-sci.org/intro/article_detail/jcm/12726.html KW - Quantization, Low bit width deep neural networks, Exact and approximate analytical formulas, Network training, Object detection. AB -

We present LBW-Net, an efficient optimization based method for quantization and training of the low bit-width convolutional neural networks (CNNs). Specifically, we quantize the weights to zero or powers of 2 by minimizing the Euclidean distance between full-precision weights and quantized weights during backpropagation (weight learning). We characterize the combinatorial nature of the low bit-width quantization problem. For 2-bit (ternary) CNNs, the quantization of $N$ weights can be done by an exact formula in $O$($N$ log $N$) complexity. When the bit-width is 3 and above, we further propose a semi-analytical thresholding scheme with a single free parameter for quantization that is computationally inexpensive. The free parameter is further determined by network retraining and object detection tests. The LBW-Net has several desirable advantages over full-precision CNNs, including considerable memory savings, energy efficiency, and faster deployment. Our experiments on PASCAL VOC dataset [5] show that compared with its 32-bit floating-point counterpart, the performance of the 6-bit LBW-Net is nearly lossless in the object detection tasks, and can even do better in real world visual scenes, while empirically enjoying more than 4× faster deployment.