/DLQoS.list

Paper list of Deep Learning QoS Papers (Accuracy vs Speed, Power, Model Size, Bitwidth, etc.)

DLQoS.list

Flexible between trade-off points (Dynamic Computation)

Power vs Accuracy

  • DrowsyNet: Convolutional neural networks with runtime power-accuracy tunability using inference-stage dropout [Paper]
    • Ren-Shuo Liu, Yun-Chen Lo, Yuan-Chun Luo, Chih-Yu Shen, and Cheng-Ju Lee, VLSI-DAT 2018.

Model size vs Accuracy

  • NestedNet: Learning Nested Sparse Structures in Deep Neural Networks [Paper]
    • Eunwoo Kim, Chanho Ahn, and Songhwai Oh, CVPR 2018.

Model Depth vs Accuracy

  • MSDNet: Multi-Scale Dense Networks for Resource Efficient Image Classification [Paper]

    • Gao Huang, Danlu Chen, Tianhong Li, Felix Wu, Laurens van der Maaten, Kilian Weinberger, ICLR 2018.

  • BranchyNet: Fast Inference via Early Exiting from Deep Neural Networks [Paper]

    • Surat Teerapittayanon, Bradley McDanel, H.T. Kung, ICPR 2016.

Un-Categorized

  • Multi-level Residual Networks from Dynamical Systems View [Paper]
    • Bo Chang, Lili Meng, Eldad Haber, Frederick Tung, David Begert, ICLR 2018.

Explore trade-off points but without flexibility

Performance (FPS) vs Accuracy

  • 3DICT: A Reliable and QoS Capable Mobile Process-In-Memory Architecture for Lookup-based CNNs in 3D XPoint ReRAMs [Paper]
    • Qian Lou, Wujie Wen, and Lei Jiang, ICCAD 2018.

Bitwidth vs Accuracy

  • WRPN: Wide Reduced-Precision Networks [Paper]
    • Asit Mishra, Eriko Nurvitadhi, Jeffrey J Cook, Debbie Marr, ICLR 2018

Depth vs Channel

  • Wide Residual Networks [Paper]
    • Sergey Zagoruyko, Nikos Komodakis, ICLR 2018