Pith. sign in

REVIEW 3 major objections 4 minor 44 references

Optimal Weighted Convolution for Classification and Denosing

T0 review · 3 major / 4 minor · 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 fixed position-dependent density mask multiplied into convolution kernels improves classification and denoising across every tested network, with no added trainable parameters.

desk verdict The paper's weighted convolution is a simple rank-1 mask, but the empirical claims are undermined by acknowledged test-set tuning of the mask parameter and by denoising baselines that are clearly broken. read the letter →

arxiv 2505.24558 v1 pith:4AXVSFPB submitted 2025-05-30 cs.CV

Simone Cammarasana , Giuseppe Patanè This is my paper
classification cs.CV
keywords ConvolutionDensityfunctionDenoisingClassificationDeeplearningweightedrank-onemaskconvolutionalneuralnetworks
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

The paper asks whether a convolution can be improved simply by remembering where each pixel sits relative to the pixel being updated. Its answer is yes: multiply each kernel element by a fixed spatial density value before applying the kernel, and every tested network improves on its target metric. The density mask is chosen as a symmetric rank-one outer product, so a $3\times3$ kernel needs one tunable scalar and a $5\times5$ kernel needs two, and it does not add any trainable parameters. On CIFAR-100 classification, VGG accuracy rises from $56.89\%$ to $66.94\%$; on DIV2K denoising, DnCNN PSNR rises from $20.17$ to $22.63$ dB. If the claim holds, the operator is a cheap, architecture-agnostic upgrade that any CNN can adopt with only a small runtime overhead.

What carries the argument

The central object is the weighted convolution $(I*W^{\Phi})^f_{ij}=\sum_{a,b}\Phi_{ab}w^f_{ab}I_{i+a-\cdots,j+b-\cdots}$ of Eq. (4): a standard discrete convolution whose kernel $w^f$ is elementwise multiplied by a density mask $\Phi$ before the inner product with the image neighbourhood. The mask is built as a symmetric rank-one outer product $\Phi=\alpha\alpha^\top$, with $\alpha$ symmetric about the center, so a $K\times K$ kernel is controlled by $(K-1)/2$ scalar hyperparameters, and setting $\alpha=(1,\dots,1)$ recovers standard convolution. The implementation precomputes $\Phi$ and forms $W^{\Phi}=\Phi\circ W$ once per iteration, adding $O(K^2F)$ operations per layer, so the operator carries its gains through a fixed spatial prior rather than through new trainable weights.

What would settle it

Re-run the same eight models with $\alpha$ selected on a held-out validation split and with the standard baselines re-implemented from their original public code, then report test metrics only for the chosen $\alpha$. If VGG's 10-point gain shrinks or disappears and DnCNN's $5\times5$ standard PSNR moves from 12.15 dB upward toward the noisy-input level, the claimed superiority would be explained by test-set peeking and a broken baseline rather than by the operator.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that the position of a pixel inside the receptive field is usable signal that standard convolution throws away, and a fixed nonuniform density can harvest it. The paper claims that for every architecture and dataset tested there exists a density matrix $\Phi=\alpha\alpha^\top$ that improves the target metric: all five CIFAR-100 classifiers gain in accuracy and F1 score, and all three DIV2K denoisers gain in PSNR and most secondary quality metrics. Because $\Phi$ multiplies existing kernel weights and is not learned, the improvement comes without increasing the number of trainable parameters; the cost is only a small runtime overhead. The method is presented as a drop-in generalization of standard convolution, which is exactly the special case $\Phi=\mathbf{1}$.

Load-bearing premise

The comparison assumes that the density values $\alpha$ reported in Tables 3 and 5 were chosen through a validation-based search without looking at the test metrics, and that the standard-convolution baselines were implemented correctly.

Editorial extensions

If this is right

  • Any existing CNN can adopt the weighted convolution as a drop-in replacement for standard convolution layers, since the number of trainable parameters is unchanged and the extra per-layer cost is only $O(K^2F)$ operations.
  • On CIFAR-100, all five classifiers improve in accuracy and F1 score, with VGG's accuracy rising from $56.89\%$ to $66.94\%$.
  • On DIV2K denoising, all three networks improve in PSNR and most quality metrics, with DnCNN's PSNR rising from $20.17$ to $22.63$ dB under a $3\times3$ kernel.
  • A weighted $5\times5$ kernel outperforms both standard and weighted $3\times3$ kernels, so the density mask can make larger receptive fields more useful without adding parameters.
  • Because uniform density recovers standard convolution, the weighted operator is a strict generalization of the standard one and can serve as the default convolution in future experiments.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper tunes $\alpha$ as a hyperparameter and never learns it; a natural next step is to backpropagate through $\alpha$ so each layer learns its own position mask, which would remove the manual search entirely.
  • Because $\Phi=\alpha\alpha^\top$ is a separable rank-one mask, the weighted convolution is equivalent to standard convolution with a kernel premultiplied by a separable profile, which suggests the same construction transfers to 3D or 1D grids by taking products of one-dimensional profiles.
  • The much larger gain for VGG than for the other classifiers suggests the practical benefit may depend on how close the standard baseline is to its full potential; an ablation with equally well-tuned baselines would clarify where the gain really comes from.
  • If the gains replicate, the density mask can be viewed as a cheap fixed spatial prior encoding that center pixels matter more, and it could be compared with learned attention or positional encodings to see what the prior captures.
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.

Referee Report

3 major / 4 minor

Summary. The paper proposes a weighted convolution operator that multiplies each kernel weight by a fixed spatial density function Phi = alpha*alpha^T (Eq. 4), with no additional trainable parameters, and evaluates it on CIFAR-100 classification and DIV2K denoising. The manuscript reports that the weighted convolution improves accuracy and PSNR/SSIM for every tested architecture compared with standard convolution, with VGG accuracy rising from 56.89% to 66.94% and DnCNN PSNR from 20.17 dB to 22.63 dB. The central claim is empirical: that this parameter-free modification of the convolution kernel consistently improves performance.

Significance. If the reported gains were obtained under sound evaluation protocols, the idea would be of practical interest because it is architecture-agnostic, adds no trainable parameters, and has a low-overhead implementation. The authors provide a public PyTorch implementation and report training times showing only modest overhead. However, the evaluation as presented does not support the central claim: the density parameter appears to have been selected on the test set, at least one baseline is implausibly broken, and no repeated-run statistics are reported. These issues make the headline improvements uninterpretable as evidence about the operator itself.

major comments (3)
  1. [Sec. 4.1, 4.2; Tables 3 and 5] The paper states in the introduction, in Section 4, and in Section 5 that the values of the density function were 'tuned to improve the accuracy' of the weighted convolution and the results of classification and denoising. No validation split, selection grid, or selection protocol is documented. Since Phi = 1 (alpha = 1) reduces the weighted convolution exactly to standard convolution, every comparison in Tables 3 and 5 is a comparison between the default alpha = 1 and a per-method, per-dataset alpha chosen after seeing performance. The reported gains (e.g., VGG 56.89% to 66.94%; DnCNN 3x3 PSNR 20.17 to 22.63 dB) are therefore consistent with test-set hyperparameter fitting, not with a general property of the operator. A proper held-out validation procedure for alpha is required before these numbers can support the paper's claim.
  2. [Table 5, DnCNN 5x5 row] The standard-convolution DnCNN baseline with a 5x5 kernel reports PSNR 12.15 dB, NRMSE 0.266, and SSIM 0.402. With Gaussian noise of sigma = 0.01 added to images in the [0,1] range, the noisy input itself has PSNR around 40 dB, so this baseline is far below the input quality and indicates a broken training or evaluation pipeline for that configuration. Because the paper's conclusion that weighted convolution 'improves the standard convolution under every metric' relies on this row, the corresponding improvement (12.15 to 23.35 dB) is not evidence for the method; it likely reflects a failed baseline rather than an improved operator.
  3. [Sec. 4.1, 4.2] All results are reported as single values with no repeated runs, seeds, or variance estimates. The claimed improvements are often a few percentage points (e.g., NiN 51.96% to 52.35%, gMLP 32.21% to 32.66%), and with the stochasticity of deep learning training these differences may not be significant. Without repeated runs or at least seed information, the reader cannot distinguish a genuine improvement from random variation, especially given the small alpha grid used.
minor comments (4)
  1. [Title and running text] The title and some running text contain the typo 'Denosing'; it should be 'Denoising'.
  2. [Fig. 1 and Tables 4, 6] The confusion-matrix figure uses 'weighed' instead of 'weighted', and Tables 4 and 6 contain the spacing artifact 'W eighted convolution'.
  3. [Sec. 4.1] The dataset name is written inconsistently as 'CIF AR-100' in the text and 'CIFAR-100' in the abstract and related work; please unify the spelling.
  4. [Sec. 3.2] The sentence 'The density function is shared across both the image and the kernels' is unclear; the density multiplies kernel weights, so a more precise statement would be that it is shared across all kernels and all spatial positions of the input.

Circularity Check

1 steps flagged · score 7.0 of 10

Reported gains are produced by tuning the density-function hyperparameters α to maximize the reported test metrics, with no documented validation split; the comparison reduces to default α=1 vs. best-of-grid α.

  1. fitted input called prediction [Section 4 (Experimental results, first paragraph); Tables 3 and 5]
    "The parallelisation and high-performance of the training allow us to tune the density function to improve the accuracy of the weighted convolution and the results of the classification (Sect. 4.1) and denoising (Sect. 4.2)."

    The weighted convolution's only new element is the density function Φ = αα^T, whose values α are treated as hyperparameters. The paper states these were tuned 'to improve the accuracy of the weighted convolution and the results', i.e., to maximize the reported metrics. Since Φ = 1 (α = 1) is exactly standard convolution, the comparison in Tables 3 and 5 is between the default α = 1 and a per-method α selected on the evaluation data. No validation-based selection protocol is given, so the reported improvements (e.g., VGG 56.89% to 66.94%; DnCNN 3×3 PSNR 20.17 to 22.63) are fitted values, not out-of-sample predictions. The claimed superiority of the operator is therefore forced by the test-set tuning of α.

full rationale

The mathematical derivation of the weighted convolution (Eq. 4) is self-contained and not circular: it reduces to standard convolution for Φ = 1, and the parameter-count analysis is straightforward. The circularity lies in the empirical evaluation. The paper explicitly says the density-function values were tuned to improve accuracy (Sec. 4, Sec. 5), and it reports results only for the tuned α without documenting a held-out validation protocol. Because α is the only thing distinguishing the method from standard convolution, the reported gains are the product of selecting α on the evaluation set rather than evidence of generalization. This matches the 'fitted input called prediction' pattern and warrants a high circularity score. Also note the DnCNN 5×5 standard baseline (PSNR 12.15 dB) is far below the noisy input level, suggesting a broken baseline that further undermines the comparison, though this is a correctness concern, not circularity. No other circular steps (self-citation, uniqueness import, ansatz smuggling, renaming) are present.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The central claim depends on density parameters α that are fitted to the experimental metrics for each model and kernel size. The rank-1 symmetric density structure is an ad hoc design choice, and the generalization of the tuned α values to unseen data is assumed without validation.

free parameters (2)
  • density parameter α1 (classification) = 0.75 (VGG), 1.15 (ResNet), 0.9 (NiN), 0.95 (gMLP), 0.8 (GAC-SNN)
    Chosen per model to improve CIFAR-100 accuracy; reported in Table 3.
  • density parameters α1, α2 (denoising) = e.g., DnCNN 3x3 α1=0.8; NAFNet 3x3 α1=0.7; DnCNN 5x5 α=(0.1,0.9)
    Chosen per model and kernel size to improve DIV2K PSNR; reported in Table 5.
assumptions (3)
  • ad hoc to paper The density function is separable and rank-1: Φ = αα^T with symmetric α; this is sufficient to improve CNN performance.
    Introduced in Section 3.2 as a design choice; no justification for rank-1 or symmetry.
  • domain assumption The density function is shared across all input channels and filters, and is fixed during training.
    Assumed in Section 3.2; the network cannot adapt the weighting to different features.
  • domain assumption The chosen α values generalize from the tuning procedure to test data.
    Required for the reported accuracy gains to be valid; the paper does not document a validation split for α.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Optimal Weighted Convolution for Classification and Denosing." pith.science (2026). https://pith.science/paper/4AXVSFPB

@misc{pith2026250524558,
  author       = {Pith},
  title        = {Pith review of: Optimal Weighted Convolution for Classification and Denosing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4AXVSFPB}},
  note         = {Machine review of arXiv:2505.24558}
}
read the original abstract

We introduce a novel weighted convolution operator that enhances traditional convolutional neural networks (CNNs) by integrating a spatial density function into the convolution operator. This extension enables the network to differentially weight neighbouring pixels based on their relative position to the reference pixel, improving spatial characterisation and feature extraction. The proposed operator maintains the same number of trainable parameters and is fully compatible with existing CNN architectures. Although developed for 2D image data, the framework is generalisable to signals on regular grids of arbitrary dimensions, such as 3D volumetric data or 1D time series. We propose an efficient implementation of the weighted convolution by pre-computing the density function and achieving execution times comparable to standard convolution layers. We evaluate our method on two deep learning tasks: image classification using the CIFAR-100 dataset [KH+09] and image denoising using the DIV2K dataset [AT17]. Experimental results with state-of-the-art classification (e.g., VGG [SZ15], ResNet [HZRS16]) and denoising (e.g., DnCNN [ZZC+17], NAFNet [CCZS22]) methods show that the weighted convolution improves performance with respect to standard convolution across different quantitative metrics. For example, VGG achieves an accuracy of 66.94% with weighted convolution versus 56.89% with standard convolution on the classification problem, while DnCNN improves the PSNR value from 20.17 to 22.63 on the denoising problem. All models were trained on the CINECA Leonardo cluster to reduce the execution time and improve the tuning of the density function values. The PyTorch implementation of the weighted convolution is publicly available at: https://github.com/cammarasana123/weightedConvolution2.0.

Figures

Figures reproduced from arXiv: 2505.24558 by the authors.

Figure 1
Figure 1. Confusion matrix: standard, weighted convolution. ResNet 56, VGG, [PITH_FULL_IMAGE:figures/full_fig_p008_1.png] view at source ↗
Figure 2
Figure 2. Denoising results with standard (second row) and weighted (third [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 3
Figure 3. Denoising results with standard (second row) and weighted (third [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

44 extracted references · 32 canonical work pages

  1. [1]

    A new convolution neural layer based on weights constraints

    Elaf Ali Abbood and Tawfiq A Al-Assadi. A new convolution neural layer based on weights constraints. In 2022 International Conference on Data Science and Intelligent Computing (ICDSIC) , pages 7--13. IEEE, 2022

    2022

  2. [2]

    Ntire 2017 challenge on single image super-resolution: Dataset and study

    Eirikur Agustsson and Radu Timofte. Ntire 2017 challenge on single image super-resolution: Dataset and study. In The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops , July 2017

    work page 2017

  3. [3]

    Noise2self: Blind denoising by self-supervision

    Joshua Batson and Loic Royer. Noise2self: Blind denoising by self-supervision. arXiv:1901.11365 , 2019

    arXiv 1901

  4. [4]

    Simple baselines for image restoration

    Liangyu Chen, Xiaojie Chu, Xiangyu Zhang, and Jian Sun. Simple baselines for image restoration. In European Conference on Computer Vision , pages 17--33. Springer, 2022

    work page 2022

  5. [5]

    Dynamic convolution: Attention over convolution kernels

    Yinpeng Chen, Xiyang Dai, Mengchen Liu, Dongdong Chen, Lu Yuan, and Zicheng Liu. Dynamic convolution: Attention over convolution kernels. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 11030--11039, 2020

    work page 2020

  6. [6]

    Convolutional kernel networks for graph-structured data

    Dexiong Chen, Laurent Jacob, and Julien Mairal. Convolutional kernel networks for graph-structured data. In International Conference on Machine Learning , pages 1576--1586. PMLR, 2020

    work page 2020

  7. [7]

    Real-time denoising of ultrasound images based on deep learning

    Simone Cammarasana, Paolo Nicolardi, and Giuseppe Patan \`e . Real-time denoising of ultrasound images based on deep learning. Medical & Biological Engineering & Computing , 60(8):2229--2244, 2022

    work page 2022

  8. [8]

    Analysis and comparison of high-performance computing solvers for minimisation problems in signal processing

    Simone Cammarasana and Giuseppe Patan \'e . Analysis and comparison of high-performance computing solvers for minimisation problems in signal processing. Mathematics and Computers in Simulation , 229:525--538, 2025

    work page 2025

Show all 44 references
  1. [9]

    Weighted convolutional neural network ensemble

    Xavier Frazao and Lu \' s A Alexandre. Weighted convolutional neural network ensemble. In Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications: 19th Iberoamerican Congress, CIARP 2014, Puerto Vallarta, Mexico, November 2-5, 2014. Proceedings 19 , ...

    2014

  2. [10]

    Deep learning for computational chemistry

    Garrett B Goh, Nathan O Hodas, and Abhinav Vishnu. Deep learning for computational chemistry. Journal of Computational Chemistry , 38(16):1291--1307, 2017

    2017

  3. [11]

    Cascadedgaze: Efficiency in global context extraction for image restoration

    Amirhosein Ghasemabadi, Muhammad Kamran Janjua, Mohammad Salameh, Chunhua Zhou, Fengyu Sun, and Di Niu. Cascadedgaze: Efficiency in global context extraction for image restoration. arXiv preprint arXiv:2401.15235 , 2024

    2024 arXiv

  4. [12]

    Generalizing the convolution operator in convolutional neural networks

    Kamaledin Ghiasi-Shirazi. Generalizing the convolution operator in convolutional neural networks. Neural Processing Letters , 50(3):2627--2646, 2019

    2019

  5. [13]

    Weighted channel dropout for regularization of deep convolutional neural network

    Saihui Hou and Zilei Wang. Weighted channel dropout for regularization of deep convolutional neural network. In Proceedings of the AAAI Conference on Artificial Intelligence , volume 33, pages 8425--8432, 2019

    2019

  6. [14]

    Delving deep into rectifiers: Surpassing human-level performance on imagenet classification

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Delving deep into rectifiers: Surpassing human-level performance on imagenet classification. In Proceedings of the IEEE international conference on computer vision , pages 1026--1034, 2015

    2015

  7. [15]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition , pages 770--778, 2016

    2016

  8. [16]

    Backpropagation and stochastic gradient descent method

    Shun ichi Amari. Backpropagation and stochastic gradient descent method. Neurocomputing , 5(4):185--196, 1993

    1993

  9. [17]

    Variable weight algorithm for convolutional neural networks and its applications to classification of seizure phases and types

    Guangyu Jia, Hak-Keung Lam, and Kaspar Althoefer. Variable weight algorithm for convolutional neural networks and its applications to classification of seizure phases and types. Pattern Recognition , 121:108226, 2022

    2022

  10. [18]

    Hyper-parameter optimization of deep learning model for prediction of parkinson’s disease

    Sukhpal Kaur, Himanshu Aggarwal, and Rinkle Rani. Hyper-parameter optimization of deep learning model for prediction of parkinson’s disease. Machine Vision and Applications , 31:1--15, 2020

    2020

  11. [19]

    Air learning: a deep reinforcement learning gym for autonomous aerial robot visual navigation

    Srivatsan Krishnan, Behzad Boroujerdian, William Fu, Aleksandra Faust, and Vijay Janapa Reddi. Air learning: a deep reinforcement learning gym for autonomous aerial robot visual navigation. Machine Learning , 110(9):2501--2540, 2021

    2021

  12. [20]

    Noise2void-learning denoising from single noisy images

    Alexander Krull, Tim-Oliver Buchholz, and Florian Jug. Noise2void-learning denoising from single noisy images. In Proc. of the IEEE Conf. on Computer Vision and Pattern Recognition , pages 2129--2137, 2019

    2019

  13. [21]

    Learning multiple layers of features from tiny images

    Alex Krizhevsky, Geoffrey Hinton, et al. Learning multiple layers of features from tiny images. 2009

    2009

  14. [22]

    Adam: A method for stochastic optimization

    Diederik P Kingma. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 , 2014

    2014 arXiv

  15. [23]

    Network in network

    Min Lin, Qiang Chen, and Shuicheng Yan. Network in network. arXiv preprint arXiv:1312.4400 , 2013

    2013 arXiv

  16. [24]

    Pay attention to mlp s

    Hanxiao Liu, Zihang Dai, David So, and Quoc V Le. Pay attention to mlp s. Advances in neural information processing systems , 34:9204--9215, 2021

    2021

  17. [25]

    Noise2noise: Learning image restoration without clean data

    Jaakko Lehtinen, Jacob Munkberg, Jon Hasselgren, Samuli Laine, Tero Karras, Miika Aittala, and Timo Aila. Noise2noise: Learning image restoration without clean data. arXiv:1803.04189 , 2018

    2018 arXiv

  18. [26]

    Omni-dimensional dynamic convolution

    Chao Li, Aojun Zhou, and Anbang Yao. Omni-dimensional dynamic convolution. In International Conference on Learning Representations , 2022

    2022

  19. [27]

    Convolutional kernel networks

    Julien Mairal, Piotr Koniusz, Zaid Harchaoui, and Cordelia Schmid. Convolutional kernel networks. Advances in Neural Information Processing Systems , 27, 2014

    2014

  20. [28]

    Gated attention coding for training high-performance and efficient spiking neural networks

    Xuerui Qiu, Rui-Jie Zhu, Yuhong Chou, Zhaorui Wang, Liang-jian Deng, and Guoqi Li. Gated attention coding for training high-performance and efficient spiking neural networks. In Proceedings of the AAAI Conference on Artificial Intelligence , volume 38, pages 601--610, 2024

    2024

  21. [29]

    Deep CNN hyperparameter optimization algorithms for sensor-based human activity recognition

    Saeid Raziani and Mehran Azimbagirad. Deep CNN hyperparameter optimization algorithms for sensor-based human activity recognition. Neuroscience Informatics , 2(3):100078, 2022

    2022

  22. [30]

    Artificial intelligence, machine learning and deep learning in advanced robotics, a review

    Mohsen Soori, Behrooz Arezoo, and Roza Dastres. Artificial intelligence, machine learning and deep learning in advanced robotics, a review. Cognitive Robotics , 3:54--70, 2023

    2023

  23. [31]

    Mobilenetv2: Inverted residuals and linear bottlenecks

    Mark Sandler, Andrew Howard, Menglong Zhu, Andrey Zhmoginov, and Liang-Chieh Chen. Mobilenetv2: Inverted residuals and linear bottlenecks. In Proceedings of the IEEE conference on computer vision and pattern recognition , pages 4510--4520, 2018

    2018

  24. [32]

    Avoiding overfitting: A survey on regularization methods for convolutional neural networks

    Claudio Filipi Gon c alves Dos Santos and Jo \ a o Paulo Papa. Avoiding overfitting: A survey on regularization methods for convolutional neural networks. ACM Computing Surveys (Csur) , 54(10s):1--25, 2022

    2022

  25. [33]

    Very deep convolutional networks for large-scale image recognition

    K Simonyan and A Zisserman. Very deep convolutional networks for large-scale image recognition. pages 1--14. Computational and Biological Learning Society, 2015

    2015

  26. [34]

    Efficientnet: Rethinking model scaling for convolutional neural networks

    Mingxing Tan and Quoc Le. Efficientnet: Rethinking model scaling for convolutional neural networks. In International Conference on Machine Learning , pages 6105--6114. PMLR, 2019

    2019

  27. [35]

    https://top500.org/lists/top500/2024/06/

    Cineca Leonardo . https://top500.org/lists/top500/2024/06/. Accessed: 2024-06-01

    2024

  28. [36]

    Attention is all you need

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, ukasz Kaiser, and Illia Polosukhin. Attention is all you need. Advances in Neural Information Processing Systems , 30, 2017

    2017

  29. [37]

    A universal image quality index

    Zhou Wang and Alan C Bovik. A universal image quality index. Signal Processing Letters , 9(3):81--84, 2002

    2002

  30. [38]

    Regularization of neural networks using dropconnect

    Li Wan, Matthew Zeiler, Sixin Zhang, Yann Le Cun, and Rob Fergus. Regularization of neural networks using dropconnect. In International conference on machine learning , pages 1058--1066. PMLR, 2013

    2013

  31. [39]

    Restormer: Efficient transformer for high-resolution image restoration

    Syed Waqas Zamir, Aditya Arora, Salman Khan, Munawar Hayat, Fahad Shahbaz Khan, and Ming-Hsuan Yang. Restormer: Efficient transformer for high-resolution image restoration. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages 5728--5739, 2022

    2022

  32. [40]

    Hyperparameter optimization in cnn for learning-centered emotion recognition for intelligent tutoring systems

    Ramon Zatarain Cabada, Hector Rodriguez Rangel, Maria Lucia Barron Estrada, and Hector Manuel Cardenas Lopez. Hyperparameter optimization in cnn for learning-centered emotion recognition for intelligent tutoring systems. Soft Computing , 24(10):7593--7602, 2020

    2020

  33. [41]

    Improvement of generalization ability of deep CNN via implicit regularization in two-stage training process

    Qinghe Zheng, Mingqiang Yang, Jiajie Yang, Qingrui Zhang, and Xinxin Zhang. Improvement of generalization ability of deep CNN via implicit regularization in two-stage training process. IEEE Access , 6:15844--15869, 2018

    2018

  34. [42]

    Beyond a gaussian denoiser: Residual learning of deep CNN for image denoising

    Kai Zhang, Wangmeng Zuo, Yunjin Chen, Deyu Meng, and Lei Zhang. Beyond a gaussian denoiser: Residual learning of deep CNN for image denoising. Transactions on Image Processing , 26(7):3142--3155, 2017

    2017

  35. [43]

    Fsim: A feature similarity index for image quality assessment

    Lin Zhang, Lei Zhang, Xuanqin Mou, and David Zhang. Fsim: A feature similarity index for image quality assessment. Transactions on Image Processing , 20(8):2378--2386, 2011

    2011

  36. [44]

    Ffdnet: Toward a fast and flexible solution for CNN -based image denoising

    Kai Zhang, Wangmeng Zuo, and Lei Zhang. Ffdnet: Toward a fast and flexible solution for CNN -based image denoising. Transactions on Image Processing , 27(9):4608--4622, 2018

    2018

Pith tools

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