[论文解读] Ternary Residual Networks
本文提出了一种三元残差网络,通过基于扰动理论指导,在敏感层添加低精度残差连接,以提升亚8比特深度神经网络的性能。该方法在使用ResNet-101时,模型大小减少1.6倍、乘法运算减少26倍,且在功耗-性能方面相比8-8比特网络提升最多2倍,仅导致约1%的精度下降,同时可通过控制残差连接实现实时精度动态调节。
Sub-8-bit representation of DNNs incur some discernible loss of accuracy despite rigorous (re)training at low-precision. Such loss of accuracy essentially makes them equivalent to a much shallower counterpart, diminishing the power of being deep networks. To address this problem of accuracy drop we introduce the notion of extit{residual networks} where we add more low-precision edges to sensitive branches of the sub-8-bit network to compensate for the lost accuracy. Further, we present a perturbation theory to identify such sensitive edges. Aided by such an elegant trade-off between accuracy and compute, the 8-2 model (8-bit activations, ternary weights), enhanced by ternary residual edges, turns out to be sophisticated enough to achieve very high accuracy ($\sim 1\%$ drop from our FP-32 baseline), despite $\sim 1.6 imes$ reduction in model size, $\sim 26 imes$ reduction in number of multiplications, and potentially $\sim 2 imes$ power-performance gain comparing to 8-8 representation, on the state-of-the-art deep network ResNet-101 pre-trained on ImageNet dataset. Moreover, depending on the varying accuracy requirements in a dynamic environment, the deployed low-precision model can be upgraded/downgraded on-the-fly by partially enabling/disabling residual connections. For example, disabling the least important residual connections in the above enhanced network, the accuracy drop is $\sim 2\%$ (from FP32), despite $\sim 1.9 imes$ reduction in model size, $\sim 32 imes$ reduction in number of multiplications, and potentially $\sim 2.3 imes$ power-performance gain comparing to 8-8 representation. Finally, all the ternary connections are sparse in nature, and the ternary residual conversion can be done in a resource-constraint setting with no low-precision (re)training.
研究动机与目标
- 解决尽管经过低精度微调,亚8比特DNN仍存在显著精度下降的问题。
- 在无需完整微调的情况下,实现在资源受限环境中的高精度、低精度推理。
- 通过启用/禁用残差连接,实现动态、实时的精度调节能力。
- 开发一种扰动理论,用于识别需针对性插入残差边的关键网络分支。
- 与8-8比特精度基线相比,实现更优的功耗-性能权衡。
提出的方法
- 该方法引入三元残差连接——在亚8比特网络的敏感分支中添加低精度连接,以补偿精度损失。
- 开发了一种扰动理论,用于识别在最小计算成本下能实现最大精度提升的层。
- 采用计算感知的容差选择方法,为各层扰动阈值(εₖ)设定,以在精度与效率之间实现最优平衡。
- 通过固定点量化(块大小N,例如N=64)构建三元残差模块,实现稀疏且高效的推理。
- 通过选择性地启用或禁用残差连接,实现在部署后对模型进行动态升级或降级。
- 从FP-32到三元残差网络的整个转换过程无需低精度微调,适用于资源受限环境的部署。
实验结果
研究问题
- RQ1通过在敏感网络分支中添加低精度残差连接,亚8比特DNN能否实现接近FP32的精度?
- RQ2如何利用扰动理论指导预训练网络中残差边插入的最佳位置选择?
- RQ3与8-8比特精度基线相比,三元残差网络在功耗-性能和模型效率方面有何提升?
- RQ4能否通过在推理时启用/禁用残差连接,实现对低精度模型精度的动态调节?
- RQ5是否可行将全精度DNN转换为三元残差网络而无需低精度微调?
主要发现
- 在ResNet-101上,8-2模型(8比特激活,三元权重)通过添加三元残差边,仅导致约1%的精度下降,接近FP32基线。
- 增强后的8-2模型使模型大小减少1.6倍,乘法运算减少26倍,且在功耗-性能方面相比8-8表示法最多提升2倍。
- 禁用最重要性最低的残差连接后,模型大小减少1.9倍,乘法运算减少32倍,精度仅比FP32下降约2%。
- 该方法可通过根据运行时精度需求动态启用或禁用残差连接,实现实时精度调节。
- 三元残差网络的转换无需低精度微调,适用于资源受限环境的部署。
- 计算感知的扰动选择在保持约1%精度损失的同时减少整体计算量,对于块大小N=64,功耗-性能增益估计为X∼5.5。
更好的研究,从现在开始
从阅读论文到最终审阅,大幅缩短您的研究时间。
无需绑定信用卡
本解读由 AI 生成,并经人工编辑审核。