ABSTRACT
Billions of sensor-enabled computing devices open tremendous opportunities for AI-powered context-aware services. Yet, democratizing AI so that heterogeneous devices can enjoy deep learning (DL) revolution requires significant reduction in computing and energy burden posed by DL models. Inspired by the concept of compressive sensing (CS), and guided by the observation that reduced sampling rates often suffice for successful classification, we devise an adaptive CS-DL pipeline. Our approach dynamically adjusts the sensing rate according to input properties and performs classification through an input-flexible DL model, demonstrating classification accuracy rates on par with uncompressed models while using up to 46% less battery energy.
- Martín Abadi, Paul Barham, Jianmin Chen, Zhifeng Chen, Andy Davis, Jeffrey Dean, Matthieu Devin, Sanjay Ghemawat, Geoffrey Irving, Michael Isard, et al. 2016. Tensorflow: A system for large-scale machine learning. In 12th {USENIX} Symposium on Operating Systems Design and Implementation ({OSDI} 16). 265--283.Google ScholarDigital Library
- Davide Anguita, Alessandro Ghio, Luca Oneto, Xavier Parra, Jorge Luis Reyes-Ortiz, et al. 2013. A public domain dataset for human activity recognition using smartphones.. In Esann, Vol. 3. 3.Google Scholar
- Konstantin Berestizshevsky and Guy Even. 2019. Dynamically sacrificing accuracy for reduced computation: Cascaded inference based on softmax confidence. In International Conference on Artificial Neural Networks. Springer, 306--320.Google ScholarDigital Library
- Tolga Bolukbasi, Joseph Wang, Ofer Dekel, and Venkatesh Saligrama. 2017. Adaptive neural networks for efficient inference. In Int. Conf. on Machine Learning (ICML). Sydney, Australia.Google Scholar
- Emmanuel J Candès, Justin Romberg, and Terence Tao. 2006. Robust uncertainty principles: Exact signal reconstruction from highly incomplete frequency information. IEEE Transactions on Information Theory 52, 2 (2006), 489--509.Google ScholarDigital Library
- Zhicheng Cui, Wenlin Chen, and Yixin Chen. 2016. Multi-scale convolutional neural networks for time series classification. arXiv preprint arXiv:1603.06995 (2016).Google Scholar
- David L Donoho. 2006. Compressed sensing. IEEE Transactions on Information Theory 52, 4 (2006), 1289--1306.Google ScholarDigital Library
- Ramin Fallahzadeh, Josue Pagan Ortiz, and Hassan Ghasemzadeh. 2017. Adaptive compressed sensing at the fingertip of internet-of-things sensors: An ultra-low power activity recognition. In Design, Automation & Test in Europe Conference & Exhibition (DATE), 2017. IEEE, 996--1001.Google ScholarCross Ref
- Xitong Gao, Yiren Zhao, Łukasz Dudziak, Robert Mullins, and Cheng-zhong Xu. 2018. Dynamic channel pruning: Feature boosting and suppression. arXiv preprint arXiv:1810.05331 (2018).Google Scholar
- Neo GPIO. 2016. A python library to control the Gpios, Accel, Gyro, Temp, Baro, Magno sensors/pins easily. Retrieved February 9th 2022 from https://github.com/smerkousdavid/Neo.GPIO/tree/master/neoGoogle Scholar
- Suyog Gupta, Ankur Agrawal, Kailash Gopalakrishnan, and Pritish Narayanan. 2015. Deep learning with limited numerical precision. In International conference on machine learning. PMLR, 1737--1746.Google Scholar
- Song Han, Jeff Pool, John Tran, and William J Dally. 2015. Learning both weights and connections for efficient neural networks. arXiv preprint arXiv:1506.02626 (2015).Google Scholar
- Andrew G Howard, Menglong Zhu, Bo Chen, Dmitry Kalenichenko, Weijun Wang, Tobias Weyand, Marco Andreetto, and Hartwig Adam. 2017. Mobilenets: Efficient convolutional neural networks for mobile vision applications. arXiv preprint arXiv:1704.04861 (2017).Google Scholar
- Michael Iliadis, Leonidas Spinoulas, and Aggelos K Katsaggelos. 2018. Deep fully-connected networks for video compressive sensing. Digital Signal Processing 72 (2018),9--18.Google ScholarCross Ref
- Monsoon Solutions Inc. 2015. Monsoon Solutions high voltage power monitor. Retrieved February 3rd 2022 from http://msoon.github.io/powermonitor/HVPM.htmlGoogle Scholar
- Sergey Ioffe and Christian Szegedy. 2015. Batch normalization: Accelerating deep network training by reducing internal covariate shift. In International conference on machine learning. PMLR, 448--456.Google Scholar
- Roua Jabla, Félix Buendía, Maha Khemaja, and Sami Faiz. 2019. Balancing Timing and Accuracy Requirements in Human Activity Recognition Mobile Applications. In Multidisciplinary Digital Publishing Institute Proceedings, Vol. 31. 15.Google Scholar
- Aftab Khan, Nils Hammerla, Sebastian Mellor, and Thomas Plötz. 2016. Optimising sampling rates for accelerometer-based human activity recognition. Pattern Recognition Letters 73 (2016), 33--40.Google ScholarDigital Library
- Timotej Knez, Octavian Machidon, and Veljko Pejović. 2021. Self-Adaptive Approximate Mobile Deep Learning. Electronics 10, 23 (2021), 2958.Google ScholarCross Ref
- Kuldeep Kulkarni, Suhas Lohit, Pavan Turaga, Ronan Kerviche, and Amit Ashok. 2016. Reconnet: Non-iterative reconstruction of images from compressively sensed measurements. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 449--458.Google ScholarCross Ref
- Kuldeep Kulkarni and Pavan Turaga. 2015. Reconstruction-free action inference from compressive imagers. IEEE transactions on pattern analysis and machine intelligence 38, 4 (2015), 772--784.Google Scholar
- Chiman Kwan, David Gribben, Akshay Rangamani, Trac Tran, Jack Zhang, and Ralph Etienne-Cummings. 2020. Detection and confirmation of multiple human targets using pixel-wise code aperture measurements. Journal of Imaging 6, 6 (2020), 40.Google ScholarCross Ref
- Ji Lin, Yongming Rao, Jiwen Lu, and Jie Zhou. 2017. Runtime neural pruning. Advances in neural information processing systems 30 (2017).Google Scholar
- Guocheng Liu, Rui Ma, and Qi Hao. 2018. A Reinforcement Learning Based Design of Compressive Sensing Systems for Human Activity Recognition. In 2018 IEEE SENSORS. IEEE, 1--4.Google Scholar
- Suhas Lohit, Kuldeep Kulkarni, Ronan Kerviche, Pavan Turaga, and Amit Ashok. 2018. Convolutional neural networks for noniterative reconstruction of compressively sensed images. IEEE Transactions on Computational Imaging 4, 3 (2018), 326--340.Google ScholarCross Ref
- Suhas Lohit, Kuldeep Kulkarni, Pavan Turaga, Jian Wang, and Aswin C Sankaranarayanan. 2015. Reconstruction-free inference on compressive measurements. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops. 16--24.Google ScholarCross Ref
- Suhas Lohit, Rajhans Singh, Kuldeep Kulkarni, and Pavan Turaga. 2018. Rate-adaptive neural networks for spatial multiplexers. arXiv preprint arXiv:1809.02850 (2018).Google Scholar
- Mohammad Malekzadeh, Richard Clegg, Andrea Cavallaro, and Hamed Haddadi. 2021. DANA: Dimension-Adaptive Neural Architecture for Multivariate Sensor Data. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 5, 3 (2021), 1--27.Google ScholarDigital Library
- Ali Mousavi and Richard G Baraniuk. 2017. Learning to invert: Signal recovery via deep convolutional networks. In 2017 IEEE international conference on acoustics, speech and signal processing (ICASSP). IEEE, 2272--2276.Google ScholarDigital Library
- Ali Mousavi, Ankit B Patel, and Richard G Baraniuk. 2015. A deep learning approach to structured signal recovery. 2015 53rd Annual Allerton Conference on Communication, Control, and Computing (Allerton) (2015), 1336--1343.Google ScholarDigital Library
- Josue Pagan, Ramin Fallahzadeh, Mahdi Pedram, Jose L Risco-Martin, Jose M Moya, Jose L Ayala, and Hassan Ghasemzadeh. 2018. Toward ultra-low-power remote health monitoring: An optimal and adaptive compressed sensing framework for activity recognition. IEEE Transactions on Mobile Computing 18, 3 (2018), 658--673.Google ScholarDigital Library
- Hamid Palangi, Rabab Ward, and Li Deng. 2016. Distributed compressive sensing: A deep learning approach. IEEE Transactions on Signal Processing 64, 17 (2016), 4504--4518.Google ScholarDigital Library
- Laurence AF Park and Simeon Simoff. 2015. Using entropy as a measure of acceptance for multi-label classification. In International symposium on intelligent data analysis. Springer, 217--228.Google ScholarCross Ref
- Nafiul Rashid, Berken Utku Demirel, and Mohammad Abdullah Al Faruque. 2021. AHAR: Adaptive CNN for Energy-efficient Human Activity Recognition in Low-power Edge Devices. arXiv preprint arXiv:2102.01875 (2021).Google Scholar
- Andreas Schuler and Gabriele Anderst-Kotsis. Houston, TX, USA, 12--14 November 2019. Examining the energy impact of sorting algorithms on Android: an empirical study. In Proceedings of the 16th EAI Int. Conf. on Mobile and Ubiquitous Systems: Computing, Networking and Services (MobiQuitous '19). ACM, 404--413.Google ScholarDigital Library
- Vanika Singhal, Angshul Majumdar, and Rabab K Ward. 2017. Semi-supervised deep blind compressed sensing for analysis and reconstruction of biomedical signals from compressive measurements. IEEE Access 6 (2017), 545--553.Google ScholarCross Ref
- Surat Teerapittayanon, Bradley McDanel, and Hsiang-Tsung Kung. 2016. Branchynet: Fast inference via early exiting from deep neural networks. In 2016 23rd International Conference on Pattern Recognition (ICPR). IEEE, 2464--2469.Google ScholarCross Ref
- Tensorflow. 2021. Pruning for on-device inference with XNNPACK. Retrieved January 24, 2022 from https://www.tensorflow.org/model_optimization/guide/pruning/pruning_for_on_device_inferenceGoogle Scholar
- UDOO. 2021. UDOO Single Board Computer. Retrieved February 3rd 2022 from https://www.udoo.org/udoo-neo/Google Scholar
- Guido Van Rossum and Fred L. Drake. 2009. Python 3 Reference Manual. CreateSpace, Scotts Valley, CA.Google Scholar
- Yibo Xu, Weidi Liu, and Kevin F Kelly. 2020. Compressed Domain Image Classification Using a Dynamic-Rate Neural Network. IEEE Access 8 (2020), 217711--217722.Google ScholarCross Ref
- Y. Yang, J. Sun, H. Li, and Z. Xu. 2020. ADMM-CSNet: A Deep Learning Approach for Image Compressive Sensing. IEEE Transactions on Pattern Analysis and Machine Intelligence 42, 3 (2020), 521--538.Google ScholarCross Ref
- Hantao Yao, Feng Dai, Shiliang Zhang, Yongdong Zhang, Qi Tian, and Changsheng Xu. 2019. DR2-net: Deep residual reconstruction network for image compressive sensing. Neurocomputing 359 (2019), 483--493.Google ScholarDigital Library
- Jiahui Yu, Linjie Yang, Ning Xu, Jianchao Yang, and Thomas Huang. 2018. Slimmable neural networks. arXiv preprint arXiv:1812.08928 (2018).Google Scholar
- Ozgur Yurur, Miguel Labrador, and Wilfrido Moreno. 2013. Adaptive and energy efficient context representation framework in mobile sensing. IEEE Transactions on Mobile Computing 13, 8 (2013), 1681--1693.Google ScholarCross Ref
- Abrar Zahin, Le Thanh Tan, and Rose Qingyang Hu. 2019. Sensor-based human activity recognition for smart healthcare: A semi-supervised machine learning. In International conference on artificial intelligence for communications and networks. Springer, 450--472.Google ScholarCross Ref
Index Terms
- Enabling resource-efficient edge intelligence with compressive sensing-based deep learning
Recommendations
Image compressive sensing via Truncated Schatten-p Norm regularization
Low-rank property as a useful image prior has attracted much attention in image processing communities. Recently, a nonlocal low-rank regularization (NLR) approach toward exploiting low-rank property has shown the state-of-the-art performance in ...
Compressive sensing via nonlocal low-rank tensor regularization
The aim of Compressing sensing (CS) is to acquire an original signal, when it is sampled at a lower rate than Nyquist rate previously. In the framework of CS, the original signal is often assumed to be sparse and correlated in some domain. Recently, ...
Image compressive sensing recovery using adaptively learned sparsifying basis via L0 minimization
From many fewer acquired measurements than suggested by the Nyquist sampling theory, compressive sensing (CS) theory demonstrates that, a signal can be reconstructed with high probability when it exhibits sparsity in some domain. Most of the ...
Comments