Towards Energy Efficient Architecture for Spaceborne Neural Networks Computation

Hybrid neural network (H-NN) has been applied in field of remote sensing. To satisfy demands of on-orbit processing that requires high throughput and restriction on power consumption, designing heterogeneous array processor becomes an effective way fulfilling various tasks. A heterogeneous array architecture is proposed to support the hybrid neural network based on characteristics of various computation types among neural network module types and of dynamic computation burden among layers. Firstly, a heterogeneous array structure consisting of different types of PEs is designed, enabling strong flexibility and high throughput. Secondly, multi-level on-chip memory structure and access strategy supporting different access modes are proposed. Thirdly, management strategy for heterogeneous computing array is designed, which combines pipelining and parallel processing to support efficient mapping of diverse hybrid neural networks. The processor has a peak throughput of up to 1.96 TOPS. The implementation on models of AlexNet, LRCN, VGG19-LSTM and CLDNN can achieve the throughput of 1.92 TOPS, 1.89 TOPS, 1.93 TOPS and 1.84 TOPS, respectively. Compared with similar neural network processor that is based on same technology, the throughput of AlexNet model is increased by 76.4%. The peak power consumption of single processor is 824 mW, to which the power restriction of on-orbit AI platform is satisfied.

[1]  John D. Evans,et al.  Improving Disaster Management Using Earth Observations—GEOSS and CEOS Activities , 2013, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing.

[2]  Hoi-Jun Yoo,et al.  14.2 DNPU: An 8.1TOPS/W reconfigurable CNN-RNN processor for general-purpose deep neural networks , 2017, 2017 IEEE International Solid-State Circuits Conference (ISSCC).

[3]  Leibo Liu,et al.  A High Energy Efficient Reconfigurable Hybrid Neural Network Processor for Deep Learning Applications , 2018, IEEE Journal of Solid-State Circuits.

[4]  Marian Verhelst,et al.  A 0.3–2.6 TOPS/W precision-scalable processor for real-time large-scale ConvNets , 2016, 2016 IEEE Symposium on VLSI Circuits (VLSI-Circuits).

[5]  Jürgen Schmidhuber,et al.  Long Short-Term Memory , 1997, Neural Computation.

[6]  Patrick T. Komiske,et al.  Deep learning in color: towards automated quark/gluon jet discrimination , 2016, Journal of High Energy Physics.

[7]  D. Tralli,et al.  Satellite remote sensing of earthquake, volcano, flood, landslide and coastal inundation hazards , 2005 .

[8]  Xiaowei Li,et al.  C-Brain: A deep learning accelerator that tames the diversity of CNNs through adaptive data-level parallelization , 2016, 2016 53nd ACM/EDAC/IEEE Design Automation Conference (DAC).

[9]  Vivienne Sze,et al.  14.5 Eyeriss: An energy-efficient reconfigurable accelerator for deep convolutional neural networks , 2016, ISSCC.

[10]  Lukasz Kaiser,et al.  One Model To Learn Them All , 2017, ArXiv.

[11]  Paris W. Vachon,et al.  Foreword to the Special Issue on Multichannel Space-Based SAR , 2015, IEEE J. Sel. Top. Appl. Earth Obs. Remote. Sens..

[12]  Dit-Yan Yeung,et al.  Convolutional LSTM Network: A Machine Learning Approach for Precipitation Nowcasting , 2015, NIPS.

[13]  Marian Verhelst,et al.  14.5 Envision: A 0.26-to-10TOPS/W subword-parallel dynamic-voltage-accuracy-frequency-scalable Convolutional Neural Network processor in 28nm FDSOI , 2017, 2017 IEEE International Solid-State Circuits Conference (ISSCC).

[14]  Nitin Chawla,et al.  14.1 A 2.9TOPS/W deep convolutional neural network SoC in FD-SOI 28nm for intelligent embedded systems , 2017, 2017 IEEE International Solid-State Circuits Conference (ISSCC).

[15]  Ronald Muellerschoen,et al.  Onboard Radar Processor Development for Rapid Response to Natural Hazards , 2016, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing.

[16]  Liangpei Zhang,et al.  Pre-Trained AlexNet Architecture with Pyramid Pooling and Supervision for High Spatial Resolution Remote Sensing Image Scene Classification , 2017, Remote. Sens..