Regularizing Activation Distribution for Training Binarized Deep Networks
Open accessThe DOI matches its Crossref record; 21 of the 21 references it cites were found in a registry or an index.
Abstract
Binarized Neural Networks (BNNs) can significantly reduce the inference latency and energy consumption in resource-constrained devices due to their pure-logical computation and fewer memory accesses. However, training BNNs is difficult since the activation flow encounters degeneration, saturation, and gradient mismatch problems. Prior work alleviates these issues by increasing activation bits and adding floating-point scaling factors, thereby sacrificing BNN's energy efficiency. In this paper, we propose to use distribution loss to explicitly regularize the activation flow, and develop a framework to systematically formulate the loss. Our experiments show that the distribution loss can consistently improve the accuracy of BNNs without losing their energy benefits. Moreover, equipped with the proposed regularization, BNN training is shown to be robust to the selection of hyper-parameters including optimizer and learning rate.
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
Adam: A Method for Stochastic Optimization
Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946
MobileNetV2: Inverted Residuals and Linear Bottlenecks
Sandler, M., Howard, A., Zhu, M., Zhmoginov, A., & Chen, L.-C. (2018) · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2018.00474 · Cited by 28,755
Identity mappings in deep residual networks
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-46493-0_38 · Cited by 10,741
Identity Mappings in Deep Residual Networks
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · arXiv · doi:10.48550/arxiv.1603.05027 · Cited by 10,665 · cited 7×
“The ResNet-18 structure uses the preactivation variant [17] with added batch normalization layers as described in Sec.”
MethodInfluentialZagoruyko, S., & Komodakis, N. (2016) · Procedings of the British Machine Vision Conference 2016 · doi:10.5244/c.30.87 · Cited by 7,244
XNOR-Net: ImageNet Classification Using Binary Convolutional Neural Networks
Rastegari, M., Ordonez, V., Redmon, J., & Farhadi, A. (2016) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-46493-0_32 · Cited by 4,408
Self-Normalizing Neural Networks
Klambauer, G., Unterthiner, T., Mayr, A., & Hochreiter, S. (2017) · arXiv · doi:10.48550/arxiv.1706.02515 · Cited by 2,267
BinaryConnect: Training Deep Neural Networks with binary weights during propagations
Courbariaux, M., Bengio, Y., & David, J.-P. (2015) · arXiv · doi:10.48550/arxiv.1511.00363 · Cited by 2,180
Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation
Bengio, Y., Léonard, N., & Courville, A. (2013) · arXiv · doi:10.48550/arxiv.1308.3432 · Cited by 1,910 · cited 1×
“, |a| < 1), then this is equivalent to using a straight-through estimator (STE) for gradient computation [3].”
MethodDoReFa-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
Hubara, I., Soudry, D., & Yaniv, R. E. (2016) · arXiv · doi:10.48550/arxiv.1602.02505 · Cited by 1,667
AMC: AutoML for Model Compression and Acceleration on Mobile Devices
He, Y., Lin, J., Liu, Z., Wang, H., Li, L.-J., & Han, S. (2018) · Lecture Notes in Computer Science · doi:10.1007/978-3-030-01234-2_48 · Cited by 1,400
Self-citationDeep Learning with Limited Numerical Precision
Gupta, S., Agrawal, A., Gopalakrishnan, K., & Narayanan, P. (2015) · arXiv · doi:10.48550/arxiv.1502.02551 · Cited by 1,364 · cited 1×
“tors to estimate the energy cost of these full-precision operations because prior quantization work shows that 16-bit fixed-point operation is generally enough for maintaining accuracy [15].”
BackgroundThe Loss Surfaces of Multilayer Networks
Choromanska, A., Henaff, M., Mathieu, M., Arous, G. B., & LeCun, Y. (2014) · arXiv · doi:10.48550/arxiv.1412.0233 · Cited by 733
Fixed Point Quantization of Deep Convolutional Networks
Lin, D. D., Talathi, S. S., & Annapureddy, V. S. (2015) · arXiv · doi:10.48550/arxiv.1511.06393 · Cited by 684 · cited 12×
“Prior work on extremely lowbit DNNs [9, 12, 28, 4, 46, 10] mainly focuses on fewbit weights and uses more bits for activations, while only a few [22, 29] target 1-bit weights and activations.”
BackgroundMethodInfluentialDeep Learning with Low Precision by Half-Wave Gaussian Quantization
Cai, Z., He, X., Sun, J., & Vasconcelos, N. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.574 · Cited by 498 · cited 14×
“use multi-level activation function for inference and variants of ReLU for gradient computation to reduce gradient mismatch [4].”
BackgroundMethodInfluentialTowards Accurate Binary Convolutional Neural Network
Lin, X., Zhao, C., & Pan, W. (2017) · arXiv · doi:10.48550/arxiv.1711.11294 · Cited by 453 · cited 4×
“approximate both weights and activations with multiple binary bases associated with floating-point coefficients [29].”
BackgroundMethodInfluentialHow to Train a Compact Binary Neural Network with High Accuracy?
Tang, W., Hua, G., & Wang, L. (2017) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v31i1.10862 · Cited by 241
WRPN: Wide Reduced-Precision Networks
Mishra, A., Nurvitadhi, E., Cook, J. J., & Marr, D. (2017) · arXiv · doi:10.48550/arxiv.1709.01134 · Cited by 227 · cited 8×
“However, reading and writing intermediate results (activations) generate a larger memory footprint than the weights [35].”
BackgroundMethodResultInfluentialSYQ: Learning Symmetric Quantization for Efficient Deep Neural Networks
Faraone, J., Fraser, N., Blott, M., & Leong, P. H. W. (2018) · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2018.00452 · Cited by 134
Research works printed but not linked: 27
- 10.1007/s11263-015-0816-y · DOI · Paper page
- rastegari, 2016, Xnornet: Imagenet classification using binary convolutional neural networks, ECCV
- maas, 2013, Rectifier nonlinearities improve neural network acoustic models, Proc ICML, 30, 3
- 10.1109/CVPR.2017.345 · DOI · Paper page
- liu, 2018, Bi-real net: Enhancing the performance of 1-bit cnns with improved representational capability and advanced training algorithm, Proceedings of the European Conference on Computer Vision (ECCV), 722
- polino, 2018, Model compression via distillation and quantization, International Conference on Learning Representations
- 10.1109/CVPR.2017.761 · DOI · Paper page
- mishra, 2018, Apprentice: Using knowledge distillation techniques to improve low-precision network accuracy, International Conference on Learning Representations
- alizadeh, 2019, A systematic study of binary neural networks’ optimisation, International Conference on Learning Representations
- 10.1109/ICCV.2017.305 · DOI
- hubara, 2017, Binarynet Pytorch
- im, 2016, An empirical analysis of deep network loss surfaces
- kawaguchi, 2016, Deep learning without poor local minima, Advances in neural information processing systems, 586
- 10.1145/3020078.3021741 · DOI · Paper page
- 10.1145/3270689 · DOI
- 10.1145/3060403.3060465 · DOI
- salimans, 2016, Weight normalization: A simple reparameterization to accelerate training of deep neural networks, Advances in neural information processing systems, 901
- glorot, 2010, Understanding the difficulty of training deep feedforward neural networks, Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics, 249
- hou, 2017, Loss-aware binarization of deep networks, International Conference on Learning Representations
- 10.1109/ICDMW.2018.00131 · DOI
- 10.1109/TCSI.2017.2698019 · DOI
- yazdani, 2018, Linear backprop in non-linear networks
- sutskever, 2013, On the importance of initialization and momentum in deep learning, International Conference on Machine Learning, 1139
- 10.1109/CVPR.2017.539 · DOI
- tieleman, 2012, Lecture 6.5-rmsprop: Divide the gradient by a running average of its recent magnitude, COURSERA Neural Networks for Machine Learning, 4, 26
- shang, 2017, Exploring normalization in deep residual networks with concatenated rectified linear units, AAAI, 1509
- 2008, CMOS065LPGP Standard Cell Library User Manual & Databook
Cited by
A Survey of Quantization Methods for Efficient Neural Network Inference
Gholami, A., Kim, S., Dong, Z., Yao, Z., Mahoney, M. W., & Keutzer, K. (2022) · Low-Power Computer Vision · doi:10.1201/9781003162810-13 · Cited by 1,219
“…block, [174] and [34] show that binarization of wider networks (i.e., networks with larger number of filters) can achieve a good trade-off between the accuracy and the model size. b) Improved Loss function: Another branch of works focuses on the choice of loss function [48, 97, 98, 243, 275].”
Binary Neural Networks: A Survey
Qin, H., Gong, R., Liu, X., Bai, X., Song, J., & Sebe, N. (2020) · Pattern Recognition · doi:10.1016/j.patcog.2020.107281 · Cited by 546
“In [85], Ding et.al. summarized the problems caused by forward binarization and backward propagation in binary neural networks, including “degeneration”, “saturation” and “gradient mismatch”.”
BackgroundA review of convolutional neural network architectures and their optimizations
Cong, S., & Zhou, Y. (2023) · Artificial Intelligence Review · doi:10.1007/s10462-022-10213-5 · Cited by 381
ReActNet: Towards Precise Binary Neural Network with Generalized Activation Functions
Liu, Z., Shen, Z., Savvides, M., & Cheng, K.-T. (2020) · Lecture Notes in Computer Science · doi:10.1007/978-3-030-58568-6_9 · Cited by 362
Self-citationForward and Backward Information Retention for Accurate Binary Neural Networks
Qin, H., Gong, R., Liu, X., Shen, M., Wei, Z., Yu, F., & Song, J. (2020) · 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr42600.2020.00232 · Cited by 337
Metadata
- Authors
- Ding, R., Chin, T.-W., Liu, Z., & Marculescu, D.
- Venue
- 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
- Publisher
- IEEE
- Published
- 2019-06
- Pages
- 11400-11409
- ISSN
- 1063-6919 (ISSN-L)
- OpenAlex
- W2982344224
- Cited by
- 151
- References
- 21
- FWCI
- 7.34
- Type
- conference paper
- Language
- en
- Open access
- green
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
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.