arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2011.01391cs.CV
收藏

参数有效的双线性投影深度神经网络

Parameter Efficient Deep Neural Networks with Bilinear Projections

Griffith University(格里菲斯大学);University of Technology Sydney(悉尼科技大学)

Litao Yu, Yongsheng Gao, Jun Zhou, Jian Zhang

摘要:

最近对深度神经网络(DNN)的研究主要集中在提高模型精度上。给定适当的深度学习框架,通常可以增加深度或层宽度以实现更高级别的准确性。然而,大量的模型参数施加更多的计算和内存使用开销,并导致参数冗余。在本文中,我们通过用双线性投影代替传统的全投影来解决DNN中的参数冗余问题。对于具有$D$输入节点和$D$输出节点的全连接层,应用双线性投影可以将模型空间复杂度从$\mathcal{O}(D^2)$降低到$\mathcal{O}(2D)$,从而实现具有次线性层大小的深度模型。然而,结构化投影具有较低的自由度相比,全投影,导致欠拟合问题。因此,我们简单地通过增加输出通道的数量来扩大映射大小,这可以保持甚至提高模型精度。这使得在内存有限的移动系统上部署这样的深度模型非常方便和高效。在四个基准数据集上的实验表明,将所提出的双线性投影应用于深度神经网络可以实现比传统的全DNN更高的精度,同时显着降低模型大小。

英文摘要:

Recent research on deep neural networks (DNNs) has primarily focused on improving the model accuracy. Given a proper deep learning framework, it is generally possible to increase the depth or layer width to achieve a higher level of accuracy. However, the huge number of model parameters imposes more computational and memory usage overhead and leads to the parameter redundancy. In this paper, we address the parameter redundancy problem in DNNs by replacing conventional full projections with bilinear projections. For a fully-connected layer with $D$ input nodes and $D$ output nodes, applying bilinear projection can reduce the model space complexity from $\mathcal{O}(D^2)$ to $\mathcal{O}(2D)$, achieving a deep model with a sub-linear layer size. However, structured projection has a lower freedom of degree compared to the full projection, causing the under-fitting problem. So we simply scale up the mapping size by increasing the number of output channels, which can keep and even boosts the model accuracy. This makes it very parameter-efficient and handy to deploy such deep models on mobile systems with memory limitations. Experiments on four benchmark datasets show that applying the proposed bilinear projection to deep neural networks can achieve even higher accuracies than conventional full DNNs, while significantly reduces the model size.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑