IoT va chekka hisoblash (Edge AI) tizimlarida neyron tarmoq modellarini optimallashtirishning apparat-dasturiy usullari
Аннотация
PDF (yuklab olish)
Mazkur ilmiy maqolada zamonaviy kompyuter injiniringi va sun’iy intellekt texnologiyalarining kesishmasida shakllangan chekka hisoblash (Edge AI) hamda Buyumlar interneti (IoT) tizimlarida chuqur o‘qitish (Deep Learning) modellarini joylashtirish va optimallashtirishning apparat-dasturiy mexanizmlari keng qamrovda tadqiq etilgan. Bulutli hisoblash (Cloud Computing) infratuzilmasiga bog‘liqlikni kamaytirish, ma’lumotlarni uzatishdagi tarmoq kechikishini (latency) bartaraf etish va axborot xavfsizligini ta’minlash zarurati resurslari cheklangan o‘rnatilgan tizimlarda (embedded systems) neyron tarmoqlarni ishga tushirishni taqozo etmoqda. Tadqiqot davomida modellarni post-trening kvantlash (PTQ), kvantlashga moslashtirilgan o‘qitish (QAT), strukturali va strukturasiz kesish (pruning), bilimlar distillyatsiyasi (Knowledge Distillation) hamda apparat tezlatgichlari (NPU, TPU, FPGA) arxitekturasiga moslashtirish usullari tahlil qilindi. Tajriba-sinov natijalari optimallashtirilgan konvolyutsion va transformer modellarining xotira hajmini 4 barobargacha qisqartirish, hisoblash tezligini 2,8–3,4 barobar oshirish hamda energiya sarfini 40% dan ortiq tejash imkonini berishini tasdiqladi.
Kalit so‘zlar: kompyuter injiniringi, sun’iy intellekt, Edge AI, Buyumlar interneti (IoT), chuqur o‘qitish, kvantlash (quantization), tarmoqni kesish (pruning), bilimlar distillyatsiyasi, neyron protsessorlar (NPU), energiya samaradorligi.
Библиографические ссылки
1. Han S., Mao H., Dally W. J. Deep Compression: Compressing Deep Neural Networks withPruning, Trained Quantization and Huffman Coding // International Conference on Learning Representations (ICLR). – 2016. – P. 1–14.
2. Jacob B., Kligys S., Chen B., Zhu M., Tang M., Howard A., Adam H., Kalenichenko D.
3. Quantization and Training of Neural Networks for Efficient Integer-Arithmetic-Only Inference // Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). – 2018. – P. 2704–2713.
4. Howard A., Sandler M., Chu G., Chen L. C., Chen B., Tan M., Wang W., Zhu Y., Pang R.,Vasudevan V., Le Q. V., Adam H. Searching for MobileNetV3 // Proceedings of the IEEE/ CVF International Conference on Computer Vision (ICCV). – 2019. – P. 1314–1324.
5. Hinton G., Vinyals O., Dean J. Distilling the Knowledge in a Neural Network // NIPS DeepLearning and Representation Learning Workshop. – 2015. – arXiv:1503.02531.
6. Lin J., Chen W. M., Lin Y., Cohn J., Gan C., Han S. MCUNet: Tiny Deep Learning on IoTDevices // Advances in Neural Information Processing Systems (NeurIPS). – 2020. – Vol.
7. 33. – P. 11711–11722.
8. Zhou Z., Chen X., Li E., Zeng L., Luo K., Zhang J. Edge Intelligence: Paving the Last Mileof Artificial Intelligence With Edge Computing // Proceedings of the IEEE. – 2019. – Vol.
9. 107, No. 8. – P. 1738–1762.
10. Gholami A., Kim S., Dong Z., Yao Z., Mahoney M. W., Keutzer K. A Survey of QuantizationMethods for Efficient Neural Network Inference // Low-Power Computer Vision. – Chapman and Hall/CRC, 2022. – P. 291–326.
11. Horowitz M. Computing's Energy Problem (and What We Can Do About It) // IEEEInternational Solid-State Circuits Conference Digest of Technical Papers (ISSCC). – 2014.
12. – P. 10–14.
13. Lai L., Suda N., Chandra V. CMSIS-NN: Efficient Neural Network Kernels for Arm CortexM CPUs // arXiv preprint arXiv:1801.06601. – 2018. – P. 1–10.
14. Shi W., Cao J., Zhang Q., Li Y., Xu L. Edge Computing: Vision and Challenges // IEEEInternet of Things Journal. – 2016. – Vol. 3, No. 5. – P. 637–646.
15. Wang X., Han Y., Leung V. C., Niyato D., Yan X., Chen X. Convergence of EdgeComputing and Deep Learning: A Comprehensive Survey // IEEE Communications Surveys & Tutorials. – 2020. – Vol. 22, No. 2. – P. 869–904.
16. Reddi V. J., Cheng C., Kanter D., Mattson P., Schmuelling G., Wu C. J. MLPerf InferenceBenchmark // Proceedings of the ACM/IEEE 47th Annual International Symposium on Computer Architecture (ISCA). – 2020. – P. 446–459.
Опубликован
Как цитировать
Выпуск
Раздел
Лицензия

Это произведение доступно по лицензии Creative Commons «Attribution» («Атрибуция») 4.0 Всемирная.

