Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 220 of the 272 references it cites were found in a registry or an index; 48 of the 149 sources its claims cite were checked for support.

A Survey of Convolutional Neural Networks: Analysis, Applications, and Prospects

Open access
Li, Zewen, Liu, Fan, Yang, Wenjie, Peng, Shouheng, Zhou, Jun · 2022 · IEEE Transactions on Neural Networks and Learning Systems

Abstract

A convolutional neural network (CNN) is one of the most significant networks in the deep learning field. Since CNN made impressive achievements in many areas, including but not limited to computer vision and natural language processing, it attracted much attention from both industry and academia in the past few years. The existing reviews mainly focus on CNN's applications in different scenarios without considering CNN from a general perspective, and some novel ideas proposed recently are not covered. In this review, we aim to provide some novel ideas and prospects in this fast-growing field. Besides, not only 2-D convolution but also 1-D and multidimensional ones are involved. First, this review introduces the history of CNN. Second, we provide an overview of various convolutions. Third, some classic and advanced CNN models are introduced; especially those key points making them reach state-of-the-art results. Fourth, through experimental analysis, we draw some conclusions and provide several rules of thumb for functions and hyperparameter selection. Fifth, the applications of 1-D, 2-D, and multidimensional convolution are covered. Finally, some open issues and promising directions for CNN are discussed as guidelines for future work.

Overview

CoChat Summary

TL;DR

This survey reviews CNN fundamentals, classic and advanced architectures, activation/loss/optimizer choices, applications across 1D/2D/multi-dimensional convolutions, and future directions. It adds comparative experiments on activation functions and optimizers, reporting that ELU often gives the best accuracy among activation functions while ReLU/Leaky ReLU are stable, and that RMSprop achieved the highest CIFAR-10 validation accuracy among tested optimizers.

Key findings

  • The survey synthesizes CNN history and models from LeNet-5 through AlexNet, VGG, GoogLeNet/Inception, ResNet, DCGAN, MobileNets, ShuffleNets, and GhostNet, emphasizing innovations such as ReLU/dropout/LRN, small kernels, residual connections, depthwise separable convolutions, channel shuffle, NAS/NetAdapt, SE, and h-swish.
  • Activation-function experiments: LeNet-5 on MNIST reached 99.20% with ELU/PReLU, 99.18% with ReLU, and 98.56% with linear; on CIFAR-10 ELU reached 65.70%, ReLU 64.40%, and linear 62.56%; on CIFAR-100 ELU reached 35.10%, Leaky ReLU 33.81%, and linear 31.24%. Pretrained VGG-16 on MNIST reached 99.55% with ReLU, 99.48% with Leaky ReLU, and 99.45% with PReLU; on CIFAR-100 Leaky ReLU reached 48.22%, PReLU 47.46%, and ReLU 44.77%, while sigmoid/tanh/ELU sometimes failed to converge.
  • Optimizer experiments with pretrained VGG-16 on CIFAR-10: RMSprop reached 87.32%, Momentum 86.37%, Nadam 86.26%, Adamax 86.18%, Adadelta 86.06%, MBGD 85.70%, Adagrad 84.68%, Nesterov 84.32%, AMSgrad 83.25%, and Adam 83.09%. Almost all tested optimizers converged, MBGD was slowest, and Nesterov/Adagrad/RMSprop/Adamax/Nadam oscillated; learning rate strongly affected convergence.
  • Applications covered include 1D CNN for time-series prediction and signal identification, 2D CNN for image classification, object detection, segmentation, and face recognition, and 3D CNN for human action recognition and object recognition/detection; future directions include model compression, security/adversarial attacks, NAS, and Capsule Neural Networks.

Methods

  • Design: a narrative survey/review plus experimental analysis, covering CNN building blocks, classic models, activation functions, loss functions, optimizers, applications, and prospects.
  • Data/sample: activation functions were tested with LeNet-5 and VGG-16 on MNIST, Fashion-MNIST, CIFAR-10, and CIFAR-100; optimizer comparison tested ten optimizers on CIFAR-10 using a pretrained VGG-16 without its last three ImageNet layers.
  • Main analysis: comparative tables/figures and rules of thumb; activation experiments used cross-entropy loss, Adam, batch sizes 256/512, and specified epochs, while optimizer experiments used cross-entropy, ReLU, batch size 512, and 100 epochs.

Limitations

  • Stated: “Due to the limitation of paper length,” the model section is an overview of representative models and emphasizes innovations “but not all details”; multi-dimensional CNN discussion uses 3D as an example because higher dimensions are uncommon and hard for humans to understand.
  • Evident: The experimental comparisons are narrow—activation functions only on LeNet-5/VGG-16 and four datasets, and optimizers only on CIFAR-10 with VGG-16; Table I reports no data augmentation, and the text does not report repeated runs or statistical significance tests.
  • Evident: The survey’s application and prospect coverage is selective rather than exhaustive, as it presents representative examples and promising directions rather than a systematic review of all CNN work.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1109/tnnls.2021.3084827

Cites

220 of 220 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

9 Valid · 37 Not valid · 67 Not checked · 41 Own results and interpretations

44 more claims · 7 with their cited sources checked

Create a free account to see all 154 claims and to verify the 52 references still open.

See all claims — free account
  • Probably not a substantial claim. 72
  • It may state two separate claims as one. 63

Create a free account to see these claims.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

5 of 5

Citations

Cites

Show all 272 references

Research works printed but not linked: 7

  1. itqan, 2016, User identification system based on finger-vein patterns using convolutional neural network, ARPN J Eng Appl Sci, 11, 3316
  2. cao, 2016, Action recognition with joints-pooled 3D deep convolutional descriptors, Proc IJCAI, 1, 3
  3. krizhevsky, 2010, Convolutional deep belief networks on CIFAR-10, Unpublished manuscript, 40, 1
  4. hagan, 2002, Neural Network Design, 9
  5. zhang, 1988, Shift-invariant pattern recognition neural network and its optical architecture, Proc Annu Conf Jpn Soc Appl Phys, 734
  6. jajodia, 2019, Image classification–cat and dog images, Image, 6, 570
  7. rapin, 2018, Nevergrad—A Gradient-Free Optimization Platform Version 0 2

Cited by

See all 5,273

Metadata

Authors
Li, Z., Liu, F., Yang, W., Peng, S., & Zhou, J.
Affiliations
  • Hohai University (CN)
  • Griffith University (AU)
Venue
IEEE Transactions on Neural Networks and Learning Systems
Published
2022-12
Volume
33
Issue
12
Pages
6999-7019
ISSN
  • 2162-237X (ISSN-L)
  • 2162-2388
OpenAlex
W3168997536
Cited by
5,273
References
272
Citations by year
2026: 853 · 2025: 1745 · 2024: 1500 · 2023: 851 · 2022: 244
FWCI
40.90
Type
journal article
Language
en
Open access
Locations
  • IEEE Transactions on Neural Networks and Learning Systems (publishedVersion)
  • PubMed (publishedVersion)
  • arXiv (Cornell University) (submittedVersion)
  • Griffith Research Online (submittedVersion)
  • arXiv (Cornell University)
Funders
  • Natural Science Foundation of Jiangsu Province (BK20191298)
  • Fundamental Research Funds for the Central Universities (B200202175)
Topics
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
  • Physical Sciences › Computer Science › Artificial Intelligence › Neural Networks and Applications
  • Physical Sciences › Computer Science › Artificial Intelligence › Domain Adaptation and Few-Shot Learning
Keywords
convolutional neural networks, multidimensional convolution, convolutional neural network architecture, hyperparameter optimization, computer vision, natural language processing, open problems
MeSH terms
Natural Language Processing, Neural Networks, Computer

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.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol and qwen3.8-omni-flash).

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 Survey of Convolutional Neural Networks: Analysis, Applications, and Prospects. https://app.cochat.ai/papers/10.1109/tnnls.2021.3084827
The paper
Li, Z., Liu, F., Yang, W., Peng, S., & Zhou, J. (2022). A Survey of Convolutional Neural Networks: Analysis, Applications, and Prospects. IEEE Transactions on Neural Networks and Learning Systems. https://doi.org/10.1109/tnnls.2021.3084827

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.