Pith. sign in

REVIEW 1 cited by

CSAGC-IDS: A Dual-Module Deep Learning Network Intrusion Detection Model for Complex and Imbalanced Data

T0 review · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A dual-module intrusion detection model using GAN oversampling plus cost-sensitive attention CNN achieves 84.55% five-class and 91.09% binary accuracy on NSL-KDD.

arxiv 2505.14027 v1 pith:OYJ2TH5G submitted 2025-05-20 cs.CR cs.AI

Yifan Zeng This is my paper · ORCID
classification cs.CRcs.AI
keywords networkcsagc-idsdetectiondataintrusionlearningmodelcomplex
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Network intrusion detection systems try to flag malicious traffic in a stream of mostly normal connections. A common problem is that some attacks, such as user-to-root (U2R) or remote-to-local (R2L), are so rare that a model never sees enough examples. This paper builds a two-part system to fix that. The first part, SC-CGAN, is a generative adversarial network that writes synthetic traffic samples for the rare classes until every class has as many examples as the most common one. It uses a self-attention mechanism to remember dependencies between features, such as the link between protocol type and port. The second part, CSCA-CNN, is a convolutional classifier that up-weights mistakes on minority classes and applies channel attention to emphasize useful feature maps. The two modules are trained separately: first the GAN synthesizes a balanced training set, then the CNN is trained on that set and tested on the untouched test set.

On the NSL-KDD benchmark, the full system reaches 84.55% accuracy and 84.52% F1 for five-class classification, and 91.09% accuracy and 92.04% F1 for binary normal-versus-attack classification. That is a small improvement over the best published results the authors compare against. The paper also applies SHAP and LIME to show which features, such as service_http or flag_REJ, push the model toward an attack label.

The main limitations are that the numbers come from a single experiment with no error bars, the code is not released, and the cost weights used to penalize rare classes are never reported. The dataset itself, NSL-KDD, is from 1999 and is considered an easy benchmark by modern standards.

Extended reading notes

Core claim

CSAGC-IDS achieves 84.55% accuracy and 84.52% F1 on five-class NSL-KDD classification and 91.09% accuracy and 92.04% F1 on binary classification, surpassing other comparative methods (Abstract, Section 4.6). The ablation study is said to 'unequivocally demonstrated the effectiveness' of the proposed enhancement (Section 4.5).

Load-bearing premise

The claimed gains of CSCA-CNN over the plain CNN (84.55% vs 80.96%) depend on the cost-sensitive weights w_i in Eq. (5), whose values are never reported and which were chosen without a described tuning procedure. If those weights were picked by looking at KDDTest+ accuracy, the improvement is a fitting artifact rather than a generalizable property; if they are arbitrary, the result is not reproducible. Additionally, the paper assumes the SC-CGAN-generated samples, validated only by downstream classifier accuracy, resemble the true KDDTest+ distribution closely enough to boost generalization.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The model rests on standard deep-learning components; no new physical or mathematical entities are introduced. The critical undisclosed quantity is the class cost weight matrix in Eq. (5), which is essential to the claimed improvement. All other listed entries are standard domain assumptions for empirical machine learning papers.

free parameters (3)
  • Cost-sensitive class weights w_i in Eq. (5) = not reported
    Central to the claimed 3.6-point gain of CSCA-CNN over CNN-only in Table 6. Values are chosen by hand and never disclosed, so the classifier is underspecified.
  • Balance target count per class = 67,343
    All classes are oversampled to the Normal class count, as shown in Table 2. The choice of the majority count as the target is a design decision that affects the training distribution.
  • GAN and classifier hyperparameters = see Table 3: lr 0.001, 5e-6, 1e-4, 0.01; batch 128; epochs 30; noise dim 123; attention dim 30; squeeze ratio 8
    Learning rates, layer counts, and bottleneck sizes are chosen by hand. No sensitivity analysis or tuning protocol is given, and baseline hyperparameters are not reported.
assumptions (4)
  • domain assumption NSL-KDD adequately represents network intrusion traffic
    Section 4.3 asserts the dataset is 'a widely recognized resource' and exhibits imbalance and complexity. The model is evaluated only on this dataset.
  • domain assumption Classifier accuracy on the balanced training set is a valid measure of generated data quality
    Section 4.4 adopts the proxy from [19]. No independent sample-quality metric (e.g., MMD, precision/recall of generated samples) is used.
  • domain assumption One-hot encoding followed by z-score standardization preserves the information needed for classification
    Section 3.1 describes preprocessing but does not verify that this representation is lossless for the downstream task.
  • domain assumption The GAN training converged to a generator whose synthetic samples match the real conditional distribution
    No training curves, mode-collapse checks, or distribution-distance metrics are reported. Figure 8 is qualitative only.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CSAGC-IDS: A Dual-Module Deep Learning Network Intrusion Detection Model for Complex and Imbalanced Data." pith.science (2026). https://pith.science/paper/OYJ2TH5G

@misc{pith2026250514027,
  author       = {Pith},
  title        = {Pith review of: CSAGC-IDS: A Dual-Module Deep Learning Network Intrusion Detection Model for Complex and Imbalanced Data},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OYJ2TH5G}},
  note         = {Machine review of arXiv:2505.14027}
}
read the original abstract

As computer networks proliferate, the gravity of network intrusions has escalated, emphasizing the criticality of network intrusion detection systems for safeguarding security. While deep learning models have exhibited promising results in intrusion detection, they face challenges in managing high-dimensional, complex traffic patterns and imbalanced data categories. This paper presents CSAGC-IDS, a network intrusion detection model based on deep learning techniques. CSAGC-IDS integrates SC-CGAN, a self-attention-enhanced convolutional conditional generative adversarial network that generates high-quality data to mitigate class imbalance. Furthermore, CSAGC-IDS integrates CSCA-CNN, a convolutional neural network enhanced through cost sensitive learning and channel attention mechanism, to extract features from complex traffic data for precise detection. Experiments conducted on the NSL-KDD dataset. CSAGC-IDS achieves an accuracy of 84.55% and an F1-score of 84.52% in five-class classification task, and an accuracy of 91.09% and an F1 score of 92.04% in binary classification task.Furthermore, this paper provides an interpretability analysis of the proposed model, using SHAP and LIME to explain the decision-making mechanisms of the model.

Figures

Figures reproduced from arXiv: 2505.14027 by the authors.

Figure 1
Figure 1. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. CSAGC-IDS process Data preprocessing is an important initial step. Numerical processing trans￾forms features into One Hot Encoding that is easily accepted by the model. That only preserves category difference information to avoid misleading the model, allowing the model to better understand the features.Normalization transforms feature values to a certain range, such as [0,1], eliminating the influence of differ￾ent… view at source ↗
Figure 3
Figure 3. SC-CGAN architecture Generative Adversarial Nets with Fusion of Conditional Information. While GANs exhibit remarkable generative capabilities, their sole reliance on noise input falls short when dealing with multi-category training data, as it lacks the ability to control the generation of specific categories. To address this lim￾itation, Conditional Generative Adversarial Networks (CGANs) [18] introduce additional… view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: CSAM architecture The integration of CSAM in SC-CGAN generators is beneficial for generating higher quality traffic data [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Residual connection of CSAM CSAM contributes to the generation of high-quality traffic data in the fol￾lowing significant ways [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: CSCA-CNN architecture Cost Sensitive learning. In ordinary classification tasks, it is often assumed that all misclassifications incur an equal cost, but in practical applications, mis￾classifying instances from different classes can lead to vastly disparate losses. To…
Figure 7
Figure 7. Figure 7: CAM architecture 4 Evaluation This section conduct extensive experiments to evaluate the proposed algorithms and model in terms of their data generation quality, classification performance, and model complexity. The results obtained validate the advantages of our pro￾p…
Figure 8
Figure 8. Figure 8: NSL-KDD original imbalanced data and SC-CGAN balanced data dimension￾ality reduction visualization by t-SNE and SAE Observing the results, it is evident that the balanced data generated by SC￾CGAN notably augments the presence of samples from rare classes compared to t…
Figure 9
Figure 9. Figure 9: 4 normal samples interpretability analysis by LIME As illustrated in [PITH_FULL_IMAGE:figures/full_fig_p015_9.png]
Figure 10
Figure 10. Figure 10: Attack and normal samples SHAP force plot By taking 100 samples and plotting force plot in horizontal stack. As illus￾trated in [PITH_FULL_IMAGE:figures/full_fig_p016_10.png]
Figure 11
Figure 11. Figure 11: 100 samples SHAP force plot 5 Conclusion CSAGC-IDS, a deep learning network intrusion detection model that leverages cost sensitive learning and a mixed attention mechanism to tackle the challenges of high-dimensional, complex, and imbalanced data distributions in net…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. $\text{C}^{2}\text{BNVAE}$: Dual-Conditional Deep Generation of Network Traffic Data for Network Intrusion Detection System Balancing

    cs.CR 2025-06 conditional novelty 4.0 of 10

    Adding class-specific batch normalization to a conditional variational autoencoder improves synthetic minority-class network traffic generation enough to nudge a Decision Tree NIDS F1-score from 72.74% to 78.19% on NSL-KDD.

Reference graph

Works this paper leans on

52 extracted references · 49 canonical work pages · cited by 1 Pith paper

  1. [1]

    Survey of intrusion detection systems: techniques, datasets and challenges

    Khraisat, A., Gondal, I., Vamplew, P., et al. Survey of intrusion detection systems: techniques, datasets and challenges. Cybersecurity,2(1), 22 (2019)

    work page 2019

  2. [2]

    FIREMAN: a toolkit for firewall modeling and analysis

    Yuan, L., Mai, J., Su, Z., et al. FIREMAN: a toolkit for firewall modeling and analysis. In: Proceedings of the 2006 IEEE Symposium on Security and Privacy. IEEE (2006)

    work page 2006

  3. [3]

    LeCun,Y.,Bengio,Y.,Hinton,G.Deeplearning.Nature, 521(7553),436–444(2015)

    work page 2015

  4. [4]

    A deep learning approach for network intrusion detection system

    Javaid, A., Niyaz, Q., Sun, W., et al. A deep learning approach for network intrusion detection system. In: Proceedings of the 9th EAI International Conference on Bio- inspired Information and Communications Technologies. pp. 21–26 (2016)

    work page 2016

  5. [5]

    He, H., Garcia, E. A. Learning from imbalanced data. IEEE Transactions on Knowl- edge & Data Engineering,20(9), 1263–1284 (2008)

    work page 2008

  6. [6]

    CSE-IDS: Using cost-sensitive deep learning and ensemble algorithms to handle class imbalance in network-based intrusion detection systems

    Gupta, N., Jindal, V., Bedi, P. CSE-IDS: Using cost-sensitive deep learning and ensemble algorithms to handle class imbalance in network-based intrusion detection systems. Computers & Security,112, 102499 (2022)

    work page 2022

  7. [7]

    A novel multi-module integrated intrusion detection system for high-dimensional imbalanced data

    Cui, J., Zong, L., Xie, J., et al. A novel multi-module integrated intrusion detection system for high-dimensional imbalanced data. Applied Intelligence,53(1), 272–288 (2023)

    work page 2023

  8. [8]

    Gradient-Based Learning Applied to Doc- ument Recognition

    LeCun, Y., Bottou, L., Bengio, Y., et al. Gradient-Based Learning Applied to Doc- ument Recognition. Proceedings of the IEEE,86(11), 2278–2324 (1998)

    work page 1998

Show all 52 references
  1. [9]

    Robust Detection for Network Intrusion of Industrial IoT Based on Multi-CNN Fusion

    Li, Y., Xu, Y., Liu, Z., et al. Robust Detection for Network Intrusion of Industrial IoT Based on Multi-CNN Fusion. Measurement,154(2), 107450 (2019)

    2019

  2. [10]

    A., Rizaner, A., Ulusoy, A

    Shams, E. A., Rizaner, A., Ulusoy, A. H. A novel context-aware feature extraction method for convolutional neural network-based intrusion detection systems. Neural Computing and Applications,33(18), 13647–13665 (2021)

    2021

  3. [11]

    A deep learning model for network intrusion detection with imbalanced data

    Fu, Y., Du, Y., Cao, Z., et al. A deep learning model for network intrusion detection with imbalanced data. Electronics,11(6), 898 (2022)

    2022

  4. [12]

    Long short-term memory

    Hochreiter, S., Schmidhuber, J. Long short-term memory. Neural Computation, 9(8), 1735–1780 (1997)

    1997

  5. [13]

    V., Bowyer, K

    Chawla, N. V., Bowyer, K. W., Hall, L. O., et al. SMOTE: Synthetic Minority Over-sampling Technique. Journal of Artificial Intelligence Research,16(1), 321– 357 (2002)

    2002

  6. [14]

    Network Intrusion Detection Combined Hybrid Sampling With Deep Hierarchical Network

    Jiang, K., Wang, W., Wang, A., et al. Network Intrusion Detection Combined Hybrid Sampling With Deep Hierarchical Network. IEEE Access,8, 32464–32476 (2020)

    2020

  7. [15]

    AESMOTE: Adversarial Reinforcement Learning With SMOTE for Anomaly Detection

    Ma, X., Shi, W. AESMOTE: Adversarial Reinforcement Learning With SMOTE for Anomaly Detection. IEEE Transactions on Network Science and Engineering, 8(2), 943–956 (2021) 18 Yifan Zeng

    2021

  8. [16]

    Generative Adversarial Nets

    Goodfellow, I., Pouget-Abadie, J., Mirza, M., et al. Generative Adversarial Nets. In: Advances in Neural Information Processing Systems 27 (NIPS 2014), Montreal, Canada (2014)

    2014

  9. [17]

    H., Park, K

    Lee, J. H., Park, K. H. GAN-based imbalanced data intrusion detection system. Personal and Ubiquitous Computing,25(1), 121–128 (2021)

    2021

  10. [18]

    Conditional Generative Adversarial Nets

    Mirza, M., Osindero, S. Conditional Generative Adversarial Nets. arXiv preprint arXiv:1411.1784 (2014)

    2014 arXiv

  11. [19]

    Effective data generation for imbalanced learning using Conditional Generative Adversarial Networks

    Douzas, G., Bacao, F. Effective data generation for imbalanced learning using Conditional Generative Adversarial Networks. Expert Systems with Applications, 82, 74–86 (2017)

    2017

  12. [20]

    Wasserstein generative adversarial networks

    Arjovsky, M., Chintala, S., Bottou, L. Wasserstein generative adversarial networks. In: Proceedings of the 34th International Conference on Machine Learning (ICML 2017), Sydney, Australia (2017)

    2017

  13. [21]

    Attention is all you need

    Vaswani, A., Shazeer, N., Parmar, N., et al. Attention is all you need. In: Proceed- ings of the 31st International Conference on Neural Information Processing Systems (2017)

    2017

  14. [22]

    B., Mann, B., Ryder, N., et al

    Brown, T. B., Mann, B., Ryder, N., et al. Language models are few-shot learners. In: Proceedings of the NeurIPS (2020)

    2020

  15. [23]

    Self-attention generative adversarial networks

    Zhang, H., Goodfellow, I., Metaxas, D., et al. Self-attention generative adversarial networks. In: Proceedings of the 36th International Conference on Machine Learning (2019)

    2019

  16. [24]

    Deep residual learning for image recognition

    He, K., Zhang, X., Ren, S., et al. Deep residual learning for image recognition. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (2016)

    2016

  17. [25]

    Y., et al

    Woo, S., Park, J., Lee, J. Y., et al. CBAM: Convolutional Block Attention Module. In: Proceedings of the 15th European Conference on Computer Vision (2018)

    2018

  18. [26]

    A Detailed Analysis of the KDD CUP 99 Data Set

    Tavallaee, M., Bagheri, E., Lu, W., et al. A Detailed Analysis of the KDD CUP 99 Data Set. In Proceedings of the 2009 IEEE Symposium on Computational Intel- ligence for Security and Defense Applications. Ottawa, ON, Canada, pp. 1-6 (2009)

    2009

  19. [27]

    P., Ba, J

    Kingma, D. P., Ba, J. Adam: A Method for Stochastic Optimization. arxiv preprint arxiv:1412.6980 (2014)

    2014 arXiv

  20. [28]

    Y., Mao, B

    Han, H., Wang, W. Y., Mao, B. H. Borderline-SMOTE: A New Over-Sampling Method in Imbalanced Data Sets Learning. In: Proceedings of the International Conference on Intelligent Computing (2005)

    2005

  21. [29]

    Improving Imbalanced Learning Through a Heuris- tic Oversampling Method Based on K-means and SMOTE

    Douzas, G., Bacao, F., Last, F. Improving Imbalanced Learning Through a Heuris- tic Oversampling Method Based on K-means and SMOTE. Information Sciences, 465, 1-20 (2018)

    2018

  22. [30]

    M., Cooper, E

    Nguyen, H. M., Cooper, E. W., Kamei, K. Borderline Over-sampling for Imbal- anced Data Classification. International Journal of Knowledge Engineering and Soft Data Paradigms (2009)

    2009

  23. [31]

    P., Welling, M., Auto-Encoding Variational Bayes

    Kingma, D. P., Welling, M., Auto-Encoding Variational Bayes. In: Proceedings of the 2nd International Conference on Learning Representations (ICLR 2014). Banff, Canada, April 14-16 (2014)

    2014

  24. [32]

    Sohn, K., Yan, X., & Lee, H. (2015). Learning Structured Output Representation using Deep Conditional Generative Models. In: Advances in Neural Information Processing Systems 28 (NIPS 2015). Montreal, Canada, December 7-12, pp. 3483- 3491 (2015)

    2015

  25. [33]

    J., Liu, B., Sheng, L., et al

    Yin, G. J., Liu, B., Sheng, L., et al. Semantics disentangling for text-to-image gen- eration. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (2019) CSAGC-IDS: A Deep Learning Network Intrusion Detection Model 19

    2019

  26. [34]

    Laurens, V. D. M., Hinton, G. Visualizing Data using t-SNE. Journal of Machine Learning Research, 9(2605): 2579-2605 (2008)

    2008

  27. [35]

    XGBoost: A Scalable Tree Boosting System

    Chen, T., Guestrin, C. XGBoost: A Scalable Tree Boosting System. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York (2016)

    2016

  28. [36]

    Intrusion detection system for NSL-KDD dataset using convo- lutional neural networks

    Ding, Y., Zhai, Y. Intrusion detection system for NSL-KDD dataset using convo- lutional neural networks. In: Proceedings of the 2018 2nd International Conference on Computer Science and Artificial Intelligence. Shenzhen, China, pp. 81–85 (2018)

    2018

  29. [37]

    F., Huang, C

    Li, Z. F., Huang, C. H., Deng, S. H., et al. A Soft Actor-Critic Reinforcement Learning Algorithm for Network Intrusion Detection. Computers & Security,135: 103502 (2023)

    2023

  30. [38]

    P., et al

    Vinayakumar, R., Alazab, M., Soman, K. P., et al. Deep learning approach for intelligent intrusion detection system. IEEE Access,7, 41525–41550 (2019)

    2019

  31. [39]

    Soft-computing-basedfalsealarmreductionforhierarchicaldataof intrusion detection system

    Singh,P.,etal."Soft-computing-basedfalsealarmreductionforhierarchicaldataof intrusion detection system." International Journal of Distributed Sensor Networks. 15(10) (2019)

    2019

  32. [40]

    Siam-IDS: Handling Class Imbalance Problem in Intrusion Detection Systems Using Siamese Neural Network

    Bedi, P., Gupta, N., Jindal, V. Siam-IDS: Handling Class Imbalance Problem in Intrusion Detection Systems Using Siamese Neural Network. In: Proceedings of the Third International Conference on Computing and Network Communications, Trivandrum (2019)

    2019

  33. [41]

    I-SiamIDS: An Improved Siam-IDS for Handling Class Imbalance in Network-Based Intrusion Detection Systems

    Bedi, P., Gupta, N., Jindal, V. I-SiamIDS: An Improved Siam-IDS for Handling Class Imbalance in Network-Based Intrusion Detection Systems. Applied Intelli- gence, 51(2): 1133-1151 (2021)

    2021

  34. [42]

    LIO-IDS: Handling Class Imbalance Using LSTM and Improved One-vs-One Technique in Intrusion Detection System

    Gupta, N., Jindal, V., Bedi, P. LIO-IDS: Handling Class Imbalance Using LSTM and Improved One-vs-One Technique in Intrusion Detection System. Computer Net- works,192: 108076 (2021)

    2021

  35. [43]

    S., Kumar, R., et al

    Sethi, K., Rupesh, E. S., Kumar, R., et al. A Context-Aware Robust Intrusion De- tection System: A Reinforcement Learning-Based Approach. International Journal of Information Security,19(6): 657-678 (2020)

    2020

  36. [44]

    Network Abnormal Traffic Detection Model Based on Semi-Supervised Deep Reinforcement Learning

    Dong, S., Xia, Y., Peng, T. Network Abnormal Traffic Detection Model Based on Semi-Supervised Deep Reinforcement Learning. IEEE Transactions on Network and Service Management, 18(4): 4197-4212 (2021)

    2021

  37. [45]

    Adversarial Environment Reinforce- ment Learning Algorithm for Intrusion Detection

    Caminero, G., Lopez-Martin, M., Carro, B. Adversarial Environment Reinforce- ment Learning Algorithm for Intrusion Detection. Computer Networks,159: 96-109 (2019)

    2019

  38. [46]

    Peeking Inside the Black-Box: A Survey on Explainable Artificial Intelligence (XAI)

    Adadi, A., Berrada, M. Peeking Inside the Black-Box: A Survey on Explainable Artificial Intelligence (XAI). IEEE Access,6, 52138-52160 (2018)

    2018

  39. [47]

    xNIDM: Explaining Deep Learning-based Net- work Intrusion Detection Systems for Active Intrusion Responses

    Wei, F., Li, H.D., Zhao, Z.M., et al. xNIDM: Explaining Deep Learning-based Net- work Intrusion Detection Systems for Active Intrusion Responses. In: Proceedings of the 32nd USENIX Security Symposium (USENIX Security 2023). Anaheim, CA, USA (2023)

    2023

  40. [48]

    Why Should I Trust You?

    Ribeiro, M.T., Singh, S., Guestrin, C. "Why Should I Trust You?" Explaining the Predictions of Any Classifier. In: Proceedings of the 22nd ACM SIGKDD In- ternational Conference on Knowledge Discovery and Data Mining. pp. 1135-1144 (2016)

    2016

  41. [49]

    A Unified Approach to Interpreting Model Predictions

    Lundberg, S.M., Lee, S.I. A Unified Approach to Interpreting Model Predictions. In: Advances in Neural Information Processing Systems 30 (2017)

    2017

  42. [50]

    A., Delas, J., Neal, C.,et al

    Merzouk, M. A., Delas, J., Neal, C.,et al. Evading deep reinforcement learning- based network intrusion detection with adversarial attacks. In: Proceedings of the 17thInternationalConferenceonAvailability,ReliabilityandSecurity,pp.1-6(2022) 20 Yifan Zeng

    2022

  43. [51]

    Distilling the knowledge in a neural network

    Hinton, G., Vinyals, O., Dean, J. Distilling the knowledge in a neural network. arxiv preprint arxiv:1503.02531 (2015)

    2015 arXiv

  44. [52]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition, pp

    Yim,J.,Joo,D.,Bae,J.,etal.Agiftfromknowledgedistillation:Fastoptimization, network minimization and transfer learning. In: Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 4133-4141 (2017)

    2017

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.