EEMS Group @ MIT (PI: Vivienne Sze)

@eems_mit

The Energy-Efficient Multimedia Systems Group develops energy-efficient systems for , &

Vrijeme pridruživanja: rujan 2016.

Tweetovi

Blokirali ste korisnika/cu @eems_mit

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @eems_mit

  1. Prikvačeni tweet

    Our paper on "Efficient Processing of Deep Neural Networks: A Tutorial and Survey" is the cover story for the December issue of Proceedings of the IEEE

    Poništi
  2. Video and slides for our talk on "Efficient Computing for Deep Learning, AI and Robotics" for MIT's Deep Learning Seminar Series are now available online!

    Poništi
  3. Enabling efficient computing is critical for the deployment of deep learning, AI and robotics. It was great to have the opportunity to discuss this in our talk for 's Deep Learning Seminar Series today.

    Poništi
  4. Exciting to see a large number of deep learning processors being developed in academia & industry! Learn how to evaluate and compare them at our tutorial on "How to Understand and Evaluate Deep Learning Processors" in San Francisco on Feb 16

    Poništi
  5. Our tutorial on "Efficient Processing of Deep Neural Networks: from Algorithms to Hardware Architectures" is now online! We discuss how to evaluate efficient approaches for processing DNNs and highlight the key questions one should ask

    Poništi
  6. proslijedio/la je Tweet

    tutorial on efficient deep learning by Vivienne Sze (). The tutorial will focus on evaluation and metrics, rather than techniques, which is a great idea!

    Prikaži ovu nit
    Poništi
  7. There's a huge number of papers on efficient processing at deep neural networks. In today's tutorial at , we discuss how to evaluate these works. Slides are

    Poništi
  8. Peter receives the Gold medal in the ACM Student Research Competition (SRC) at MICRO-52 for his work on “A Mutual Information Accelerator for Autonomous Robot Exploration” (an extension of the RSS2019 paper). Check out his poster at

    Prikaži ovu nit
    Poništi
  9. If you are , and interested in learning about fast energy estimation for designing and evaluating domain-specific accelerators (e.g., Eyeriss for deep neural networks), we invite you to check out Nellie's talk on Accelergy this afternoon in session 7B at !

    Poništi
  10. We will be giving a tutorial on "Efficient Processing of Deep Neural Network: from Algorithms to Hardware Architectures" at (Dec. 9). We will discuss methods to enable efficient processing of DNNs across the entire stack.

    Poništi
  11. If you are attending , and interested in learning about tools for evaluating deep neural network accelerator designs, check out the timeloop/accelergy tutorial this afternoon!

    Poništi
  12. Prikaži ovu nit
    Poništi
  13. Full house at tutorial on Efficient Image Processing with Deep Neural Networks. We covered techniques (eg. network architecture design, neural architecture search, designing w/ hardware in the loop) & applications (eg. depth est, seg, super-res).

    Poništi
  14. Designing efficient DNNs is critical for AI deployment on mobile devices. NetAdapt automatically adapts & simplifies DNNs for a platform w/ few parameters to tune for ease of use. It was used for MobileNetV3 & FastDepth. Check out the code @

    Poništi
  15. If you are attending and interested in learning about tools for evaluating deep neural network accelerator designs, we invite you to check out our tutorial on Acclergy and Timeloop taught by . More info at

    Poništi
  16. How can Agile and Open Hardware accelerate our design process for applications such as AI, Robotics and Video Compression? Check out our talk at the SIGARCH Visioning Workshop @

    Poništi
  17. We will be giving a tutorial on "Efficient Image Processing with Deep Neural Networks" at the in Tapei (Sept 22). We will discuss applications such as image classification, depth estimation, image segmentation, and super-resolution. Register @

    Prikaži ovu nit
    Poništi
  18. Estimating energy consumption is a key step in designing domain-specific accelerators. Our latest work, Accelergy, performs energy estimations without requiring a complete hardware description allowing the fast exploration of the accelerator design space:

    Poništi
  19. It was great to be part of the inaugural TEDxMIT! To learn more about how to bring AI to the palm of your hand to tackle challenges in robotics and health care, check out the talk at

    Poništi
  20. Accelerating Shannon Mutual Information is critical for high-speed robot exploration. Our high-throughput accelerator is two orders of magnitude faster than existing solutions @ >10x lower power. Check out our paper Joint work with

    Prikaži ovu nit
    Poništi
  21. Doing exciting research at the intersection of Systems and Machine Learning? Submit your work to the Conference of Systems and Machine Learning. Deadline Sept 9! CFP at Check out videos from previous iterations SysML18, SysML19 at

    Poništi

Čini se da učitavanje traje već neko vrijeme.

Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.

    Možda bi vam se svidjelo i ovo:

    ·