---
_id: '18247'
abstract:
- lang: eng
  text: Convolutional neural networks (CNNs) achieve state-of-the-art accuracy in
    a variety of tasks in computer vision and beyond. One of the major obstacles hindering
    the ubiquitous use of CNNs for inference on low-power edge devices is their high
    computational complexity and memory bandwidth requirements. The latter often dominates
    the energy footprint on modern hardware. In this paper, we introduce a lossy transform
    coding approach, inspired by image and video compression, designed to reduce the
    memory bandwidth due to the storage of intermediate activation calculation results.
    Our method does not require fine-tuning the network weights and halves the data
    transfer volumes to the main memory by compressing feature maps, which are highly
    correlated, with variable length coding. Our method outperform previous approach
    in term of the number of bits per value with minor accuracy degradation on ResNet-34
    and MobileNetV2. We analyze the performance of our approach on a variety of CNN
    architectures and demonstrate that FPGA implementation of ResNet-18 with our approach
    results in a reduction of around 40% in the memory energy footprint, compared
    to quantized network, with negligible impact on accuracy. When allowing accuracy
    degradation of up to 2%, the reduction of 60% is achieved. A reference implementation
    accompanies the paper.
article_number: '9206968'
article_processing_charge: No
arxiv: 1
author:
- first_name: Brian
  full_name: Chmiel, Brian
  last_name: Chmiel
- first_name: Chaim
  full_name: Baskin, Chaim
  last_name: Baskin
- first_name: Evgenii
  full_name: Zheltonozhskii, Evgenii
  last_name: Zheltonozhskii
- first_name: Ron
  full_name: Banner, Ron
  last_name: Banner
- first_name: Yevgeny
  full_name: Yermolin, Yevgeny
  last_name: Yermolin
- first_name: Alex
  full_name: Karbachevsky, Alex
  last_name: Karbachevsky
- first_name: Alexander
  full_name: Bronstein, Alexander
  id: 58f3726e-7cba-11ef-ad8b-e6e8cb3904e6
  last_name: Bronstein
  orcid: 0000-0001-9699-8730
- first_name: Avi
  full_name: Mendelson, Avi
  last_name: Mendelson
citation:
  ama: 'Chmiel B, Baskin C, Zheltonozhskii E, et al. Feature map transform coding
    for energy-efficient CNN inference. In: <i>2020 International Joint Conference
    on Neural Networks (IJCNN)</i>. IEEE; 2020. doi:<a href="https://doi.org/10.1109/ijcnn48605.2020.9206968">10.1109/ijcnn48605.2020.9206968</a>'
  apa: 'Chmiel, B., Baskin, C., Zheltonozhskii, E., Banner, R., Yermolin, Y., Karbachevsky,
    A., … Mendelson, A. (2020). Feature map transform coding for energy-efficient
    CNN inference. In <i>2020 International Joint Conference on Neural Networks (IJCNN)</i>.
    Glasgow, United Kingdom: IEEE. <a href="https://doi.org/10.1109/ijcnn48605.2020.9206968">https://doi.org/10.1109/ijcnn48605.2020.9206968</a>'
  chicago: Chmiel, Brian, Chaim Baskin, Evgenii Zheltonozhskii, Ron Banner, Yevgeny
    Yermolin, Alex Karbachevsky, Alex M. Bronstein, and Avi Mendelson. “Feature Map
    Transform Coding for Energy-Efficient CNN Inference.” In <i>2020 International
    Joint Conference on Neural Networks (IJCNN)</i>. IEEE, 2020. <a href="https://doi.org/10.1109/ijcnn48605.2020.9206968">https://doi.org/10.1109/ijcnn48605.2020.9206968</a>.
  ieee: B. Chmiel <i>et al.</i>, “Feature map transform coding for energy-efficient
    CNN inference,” in <i>2020 International Joint Conference on Neural Networks (IJCNN)</i>,
    Glasgow, United Kingdom, 2020.
  ista: Chmiel B, Baskin C, Zheltonozhskii E, Banner R, Yermolin Y, Karbachevsky A,
    Bronstein AM, Mendelson A. 2020. Feature map transform coding for energy-efficient
    CNN inference. 2020 International Joint Conference on Neural Networks (IJCNN).
    International Joint Conference on Neural Networks, 9206968.
  mla: Chmiel, Brian, et al. “Feature Map Transform Coding for Energy-Efficient CNN
    Inference.” <i>2020 International Joint Conference on Neural Networks (IJCNN)</i>,
    9206968, IEEE, 2020, doi:<a href="https://doi.org/10.1109/ijcnn48605.2020.9206968">10.1109/ijcnn48605.2020.9206968</a>.
  short: B. Chmiel, C. Baskin, E. Zheltonozhskii, R. Banner, Y. Yermolin, A. Karbachevsky,
    A.M. Bronstein, A. Mendelson, in:, 2020 International Joint Conference on Neural
    Networks (IJCNN), IEEE, 2020.
conference:
  end_date: 2020-07-24
  location: Glasgow, United Kingdom
  name: International Joint Conference on Neural Networks
  start_date: 2020-07-19
date_created: 2024-10-08T13:04:52Z
date_published: 2020-09-28T00:00:00Z
date_updated: 2024-12-12T10:04:54Z
day: '28'
doi: 10.1109/ijcnn48605.2020.9206968
extern: '1'
external_id:
  arxiv:
  - '1905.10830'
language:
- iso: eng
main_file_link:
- open_access: '1'
  url: https://doi.org/10.48550/arXiv.1905.10830
month: '09'
oa: 1
oa_version: Preprint
publication: 2020 International Joint Conference on Neural Networks (IJCNN)
publication_identifier:
  eissn:
  - 2161-4407
  isbn:
  - '9781728169279'
publication_status: published
publisher: IEEE
quality_controlled: '1'
scopus_import: '1'
status: public
title: Feature map transform coding for energy-efficient CNN inference
type: conference
user_id: 3E5EF7F0-F248-11E8-B48F-1D18A9856A87
year: '2020'
...