A Network-Centric Hardware/Algorithm Co-Design to Accelerate Distributed Training of Deep Neural Networks
The DOI matches its Crossref record; 11 of the 11 references it cites were found in a registry or an index.
Abstract
Training real-world Deep Neural Networks (DNNs) can take an eon (i.e., weeks or months) without leveraging distributed systems. Even distributed training takes inordinate time, of which a large fraction is spent in communicating weights and gradients over the network. State-of-the-art distributed training algorithms use a hierarchy of worker-aggregator nodes. The aggregators repeatedly receive gradient updates from their allocated group of the workers, and send back the updated weights. This paper sets out to reduce this significant communication cost by embedding data compression accelerators in the Network Interface Cards (NICs). To maximize the benefits of in-network acceleration, the proposed solution, named INCEPTIONN (In-Network Computing to Exchange and Process Training Information Of Neural Networks), uniquely combines hardware and algorithmic innovations by exploiting the following three observations. (1) Gradients are significantly more tolerant to precision loss than weights and as such lend themselves better to aggressive compression without the need for the complex mechanisms to avert any loss. (2) The existing training algorithms only communicate gradients in one leg of the communication, which reduces the opportunities for in-network acceleration of compression. (3) The aggregators can become a bottleneck with compression as they need to compress/decompress multiple streams from their allocated worker group. To this end, we first propose a lightweight and hardware-friendly lossy-compression algorithm for floating-point gradients, which exploits their unique value characteristics. This compression not only enables significantly reducing the gradient communication with practically no loss of accuracy, but also comes with low complexity for direct implementation as a hardware block in the NIC. To maximize the opportunities for compression and avoid the bottleneck at aggregators, we also propose an aggregator-free training algorithm that exchanges gradients in both legs of communication in the group, while the workers collectively perform the aggregation in a distributed manner. Without changing the mathematics of training, this algorithm leverages the associative property of the aggregation operator and enables our in-network accelerators to (1) apply compression for all communications, and (2) prevent the aggregator nodes from becoming bottlenecks. Our experiments demonstrate that INCEPTIONN reduces the communication time by 70.9~80.7% and offers 2.2~3.1x speedup over the conventional training system, while achieving the same level of accuracy.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474
ImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155
LeCun, Y., Bengio, Y., & Hinton, G. (2015) · Nature · doi:10.1038/nature14539 · Cited by 84,126
Gradient-based learning applied to document recognition
Lecun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998) · Proceedings of the IEEE · doi:10.1109/5.726791 · Cited by 58,805
In-Datacenter Performance Analysis of a Tensor Processing Unit
Jouppi, N. P., Young, C., Patil, N., Patterson, D., Agrawal, G., Bajwa, R., Bates, S., Bhatia, S., Boden, N., Borchers, A., et al. (2017) · Proceedings of the 44th Annual International Symposium on Computer Architecture · doi:10.1145/3079856.3080246 · Cited by 4,537 · cited 4×
“Google proposes the TPU [22], which is an accelerator with the systolic array architecture for the inference of neural networks.”
BackgroundInfluentialAccurate, Large Minibatch SGD: Training ImageNet in 1 Hour
Goyal, P., Dollár, P., Girshick, R., Noordhuis, P., Wesolowski, L., Kyrola, A., Tulloch, A., Jia, Y., & He, K. (2017) · arXiv · doi:10.48550/arxiv.1706.02677 · Cited by 2,294 · cited 1×
“if (t=NO COMPRESS) then f ←v[31:0] else if (t=0BIT COMPRESS) then f ←32’b0 else if (t=8BIT COMPRESS) then s←v[7] n shi f t← f irst1 loc f rom MSB (v[6:0]) m←concat(v[6:0]<<n shi f t, 16’b0) else if (t=16BIT COMPRESS) then s←v[15] n shi f t← f irst1 loc f rom MSB (v[14:0]) m←concat(v[14:0]<<n shi f t, 8’b0) end e←127−n…”
BackgroundDoReFa-Net: Training Low Bitwidth Convolutional Neural Networks with Low Bitwidth Gradients
Zhou, S., Wu, Y., Ni, Z., Zhou, X., Wen, H., & Zou, Y. (2016) · arXiv · doi:10.48550/arxiv.1606.06160 · Cited by 1,690
Hogwild: A Lock-Free Approach to Parallelizing Stochastic Gradient Descent
Niu, F., Recht, B., Re, C., & Wright, S. J. (2011) · arXiv · doi:10.48550/arxiv.1106.5730 · Cited by 1,211 · cited 1×
“Similar to the aforementioned algorithms, HogWild! [80], DistBelief [1] and SSP [81] also take the worker-aggregator approach, while they perform asynchronous update of gradients during training to reduce synchronization overhead.”
Eyeriss: a spatial architecture for energy-efficient dataflow for convolutional neural networks
Chen, Y.-H., Emer, J., & Sze, V. (2016) · ACM SIGARCH Computer Architecture News · doi:10.1145/3007787.3001177 · Cited by 1,182 · cited 1×
“Most of the work have concentrated on the inference phase [22, 54–70] while INCEPTIONN specifically aims for accelerating the training phase.”
MethodDeep Gradient Compression: Reducing the Communication Bandwidth for Distributed Training
Lin, Y., Han, S., Mao, H., Wang, Y., & Dally, W. J. (2017) · arXiv · doi:10.48550/arxiv.1712.01887 · Cited by 1,134
Cited by
Edge Intelligence: Paving the Last Mile of Artificial Intelligence With Edge Computing
Zhou, Z., Chen, X., Li, E., Zeng, L., Luo, K., & Zhang, J. (2019) · Proceedings of the IEEE · doi:10.1109/jproc.2019.2918951 · Cited by 2,455
“INCEPTIONN [68] Hybrid Level-5 • Maximizing the opportunities for”
BackgroundDeep Learning With Edge Computing: A Review
Chen, J., & Ran, X. (2019) · Proceedings of the IEEE · doi:10.1109/jproc.2019.2921977 · Cited by 1,540
Edge Intelligence in Intelligent Transportation Systems: A Survey
Gong, T., Zhu, L., Yu, F. R., & Tang, T. (2023) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2023.3275741 · Cited by 256
Zhou, Z., Yang, S., Pu, L., & Yu, S. (2020) · IEEE Internet of Things Journal · doi:10.1109/jiot.2020.2984332 · Cited by 77
“Along a different route, Li et al. [16] demonstrated how to quantizing the model updates to low-precision float-value representations to reduce the communication overhead.”
Background
Metadata
- Authors
- Li, Y., Park, J., Alian, M., Yuan, Y., Qu, Z., Pan, P., Wang, R., Schwing, A., Esmaeilzadeh, H., & Kim, N. S.
- Venue
- 2018 51st Annual IEEE/ACM International Symposium on Microarchitecture (MICRO)
- Publisher
- IEEE
- Published
- 2018-10
- Pages
- 175-188
- ISSN
- 1072-4451 (ISSN-L)
- OpenAlex
- W2905104204
- Cited by
- 85
- References
- 11
- FWCI
- 11.32
- Type
- conference paper
- Language
- en
- Open access
- closed
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: A Network-Centric Hardware/Algorithm Co-Design to Accelerate Distributed Training of Deep Neural Networks. https://app.cochat.ai/papers/10.1109/micro.2018.00023
- The paper
- Li, Y., Park, J., Alian, M., Yuan, Y., Qu, Z., Pan, P., Wang, R., Schwing, A., Esmaeilzadeh, H., & Kim, N. S. (2018). A Network-Centric Hardware/Algorithm Co-Design to Accelerate Distributed Training of Deep Neural Networks. 2018 51st Annual IEEE/ACM International Symposium on Microarchitecture (MICRO). https://doi.org/10.1109/micro.2018.00023
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.