access icon free Energy-efficient and reliable in-memory classifier for machine-learning applications

© The Institution of Engineering and Technology
Received 04/02/2019, Accepted 08/05/2019, Revised 25/04/2019, Published 11/06/2019

Large-scale machine-learning (ML) algorithms require extensive memory interactions. Managing or reducing data movement can significantly increase the speed and efficiency of many ML tasks. Towards this end, the authors devise an energy efficient in-memory computing (IMC) kernel for linear classification and design an initial prototype. The authors achieve a power savings of over 6.4 times than a conventional discrete system while improving reliability by 54.67%. The authors employ a split-data-aware technique to manage process, voltage, and temperature variations and to achieve fair trade-offs between energy efficiency, area requirements, and accuracy. The authors utilise a trimodal architecture with a hierarchical tree structure to further decrease power consumption. The authors also explore alternatives to the hierarchical tree structure with a significantly reduced number of linear regression blocks, while maintaining a competitive classification accuracy. Overall, the scheme provides a fast, energy efficient, and competitively accurate binary classification kernel.

Inspec keywords: trees (mathematics); buffer storage; learning (artificial intelligence)

Other keywords: large-scale machine-learning algorithms; data movement; in-memory computing kernel; split-data-aware technique; binary classification kernel; linear classification; energy-efficient- reliable in-memory classifier; energy efficiency; ML tasks; hierarchical tree structure; extensive memory interactions; area requirements

Subjects: Semiconductor storage; Knowledge engineering techniques; Combinatorial mathematics; Combinatorial mathematics; Memory circuits

References

    1. 1)
      • 7. Agrawal, A., Jaiswal, A., Lee, C., et al: ‘X-SRAM: enabling in-memory Boolean computations in CMOS static random access memories’, IEEE Trans. Circuits Syst. I, 2018, 65, (12), pp. 4219–4232.
    2. 2)
      • 30. Lee, K.T., Kang, C.Y., Yoo, O.S., et al: ‘PBTI-associated high-temperature hot carrier degeneration of nMOSFETs with metal-gate/high-k dielectrics’, IEEE Electron Device Lett., 2008, 29, (4), pp. 389–391.
    3. 3)
      • 18. Park, S.P., Kang, K., Roy, K.: ‘Reliability implications of bias-temperature instability in digital ICs’, IEEE Des. Test Comput., 2009, 26, (6), pp. 8–17.
    4. 4)
      • 8. Dong, Q., Jeloka, S., Saligane, M., et al: ‘A 4+2T SRAM for searching and in-memory computing with 0.3-V VDDmin’, IEEE J. Solid-State Circuits, 2018, 53, pp. 1006–1015.
    5. 5)
      • 14. Gong, N., Jiang, S., Wang, J., et al: ‘Hybrid-cell register files design for improving NBTI reliability’, Microelectron. Reliab., 2012, 52, (9-10), pp. 1865–1869.
    6. 6)
      • 16. Ahmed, F., Milor, L.: ‘NBTI resistant SRAM design’. 4th IEEE Int. Workshop on Advances in Sensors and Interfaces (IWASI), 2011, Savelletri di Fasano, Italy, 2011.
    7. 7)
      • 6. Gao, M., Ayers, G., Kozyrakis, C.: ‘Practical near-data processing for in-memory analytics frameworks’. 2015 Int. Conf. on Parallel Architecture and Compilation (PACT), Washington, DC, USA, 2015.
    8. 8)
      • 17. Bansal, B., Rahul, R., Jae-Joon, K., et al: ‘Impacts of NBTI and PBTI on SRAM static/dynamic noise margins and cell failure probability’, Microelectron. Reliab., 2009, 49, (6), pp. 642–649.
    9. 9)
      • 11. Kang, M., Gonugondla, S.K., Patil, A., et al: ‘A multi-functional in-memory inference processor using a standard 6T SRAM array’, IEEE J. Solid-State Circuits, 2018, 53, (2), pp. 642–655.
    10. 10)
      • 4. Steinhardt, J., Duchi, J.: ‘Minimax rates for memory-bounded sparse linear regression’. Proc. of the 28th Conf. on Learning Theory, Paris, France, 2015.
    11. 11)
      • 25. Zhang, J., Wang, Z., Verma, N.: ‘In-memory computation of a machine-learning classifier in a standard 6T SRAM array’, J. Solid-State Circuits, 2017, 52, (4), pp. 915–924.
    12. 12)
      • 27. Kim, J.K., Knag, P., Chen, T., et al: ‘A 640 m pixel/s 3.65 mW sparse event-driven neuromorphic object recognition processor with onn-chip learning’. IEEE Symp. on VLSI Circuits, Kyoto, Japan, 2015.
    13. 13)
      • 28. Kang, M., Lim, S., Gonugondla, S., et al: ‘An in-memory VLSI architecture for convolutional neural networks’, J. Emerg. Sel. Top. Circuits Syst., 2018, 8, (3), pp. 494–505.
    14. 14)
      • 22. Yabuuchi, M., Kobayashi, K.: ‘Circuit characteristic analysis considering NBTI and PBTI-induced delay degradation’. IEEE Int. Meting For Future of Electron Devices, Kansai, 2012.
    15. 15)
      • 29. Moons, B., Verhelst, M.: ‘A 0.3–2.6 TOPS/w precision-scalable processor for real-time large-scale ConvNets’. IEEE Symp. on VLSI Circuits, Honolulu, HI, USA, 2016.
    16. 16)
      • 26. Jiang, S., Priya, S.R., Elango, N., et al: ‘An energy-efficient in-memory computing machine learning classifier scheme’. Int. Conf. on VLSI Design, Delhi, India, 2019.
    17. 17)
      • 19. Sanjay, V., Chris, H., Sapatnekar, S.: ‘Impact of NBTI on SRAM read stability and design for reliability’. Int. Symp. on Quality Electronic Design, San Jose, CA, USA, 2006.
    18. 18)
      • 24. Chang, I.J., Mohapatra, D., Roy, K.: ‘A priority-based 6T/8T hybrid SRAM architecture for aggressive voltage scaling in video applications’, IEEE Trans. Circuits Syst. Video Technol., 2011, 21, (2), pp. 101–112.
    19. 19)
      • 3. Cunningham, J.P., Ghahramani, Z.: ‘Linear dimensionality reduction: survey, insights, and generalizations’, J. Mach. Learn. Res., 2015, 16, pp. 2859–2900.
    20. 20)
      • 12. Kang, M., Gonugondle, S., Shanbag, N.: ‘A 19.4 nJ/decision 364 k decisions/s in-memory random forest classifier in 6T SRAM array’. IEEE European Solid State Circuits Conf., Leuven, Belgium, 2017.
    21. 21)
      • 21. Zhao, W., Cao, Y.: ‘New generation of predictive technology model for subub-45 nm early design exploration’, IEEE Trans. Electron Devices, 2006, 53, (11), pp. 2816–2823.
    22. 22)
      • 5. Horowitz, M.: ‘1.1 computing's energy problem (and what we can do about it)’. 2014 IEEE Int. Solid-State Circuits Conf. Digest of Technical Papers (ISSCC), San Francisco, CA, USA, 2014.
    23. 23)
      • 10. Kang, M., Shanbhag, N.R.: ‘In-memory computing architectures for sparse distributed memory’, IEEE Trans. Biomed. Circuits Syst., 2016, 10, (4), pp. 855–863.
    24. 24)
      • 1. Shen, F., Shen, C., Liu, W., et al: ‘Supervised discrete hashing’. Proc. of the IEEE Conf. on Computer Vision and Pattern Recognition, Boston, 2015.
    25. 25)
      • 9. Yu, H., Wang, Y., Chen, S., et al: ‘Energy efficient in-memory machine learning for data intensive image processing by non-volatile domain-wall memory’. Asia and South Pacific Design and Automation Conf., Suntec City, Singapore, 2014.
    26. 26)
      • 13. ‘The MNIST database of handwritten digits’. Available at http://yann.lecun.com/exdb/mnist/, accessed June 2018.
    27. 27)
      • 2. Huang, H.Y., Lin, C.J.: ‘Linear and kernel classification: when to use which?’. Proc. of the 2016 SIAM Int. Conf. on Data Mining, Miami, FL, USA, 2016.
    28. 28)
      • 15. Vattikonda, R., Wang, W., Cao, Y.: ‘Modeling and minimization of PMOS NBTI effect for robust nanometer design’. Design Automation Conf., San Francisco, 2006.
    29. 29)
      • 20. Bhardwaj, S., Wang, W., Vattikonda, R., et al: ‘Predictive modeling of the NBTI effect for reliable design’. Custom Integrated Circuits Conf., San Jose, CA, USA, 2006.
    30. 30)
      • 23. Bairy, B., Craig, T., Eshaghian-Wilner, M., et al: ‘Mitigating the impact of NBTI and PBTI degradation’, Global J. Technol. Optim., 2016, 7, (2), p. 195.
    31. 31)
      • 31. Wang, W., Reddy, V., Yang, B., et al: ‘Statistical prediction of circuit aging under process variations’. Custom Integrated Circuits Conf., San Jose, CA, USA, 2008.
http://iet.metastore.ingenta.com/content/journals/10.1049/iet-cdt.2019.0040
Loading

Related content

content/journals/10.1049/iet-cdt.2019.0040
pub_keyword,iet_inspecKeyword,pub_concept
6
6
Loading