Pith. sign in

REVIEW 4 major objections 4 minor 44 references

Optimal Density Functions for Weighted Convolution in Learning Models

T0 review · 4 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 The paper claims that scaling convolution kernels by a learned density function — one weight per kernel position, shared across the network — reduces a denoising network's loss by up to 53% and raises a classifier's accuracy from 46% to…

desk verdict A clearly written extension of convolution with a trainable symmetric mask, but the headline accuracy gains are not backed by held-out evaluation. read the letter →

arxiv 2505.24527 v1 pith:KPQ7YFVR submitted 2025-05-30 cs.CV cs.LG

Simone Cammarasana , Giuseppe Patanè This is my paper · ORCID
classification cs.CVcs.LG MSC 68T0765K10
keywords weightedconvolutionoptimaldensityfunctionoptimisationmodeldeeplearningimagedenoisingDIRECT-Lconvolutionalneuralnetwork
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

Standard convolution treats every neighbourhood offset equally; this paper proposes instead to multiply each kernel position by a density value that controls how much that offset contributes, and to learn those density values from data. The paper's central finding is that the optimal density is not uniform: for a 5×5 kernel the density vector $\alpha = [0.38, 2.21, 1, 2.21, 0.38]$ reduces the denoising loss by 53% compared with uniform $\alpha = [1, 1, 1, 1, 1]$, with 12% and 30% reductions for 3×3 and 7×7 kernels. The density is shared across kernels and images, symmetric, and built as the rank-one outer product $\Phi = \alpha \alpha^{\top}$, so a kernel of size $K$ adds only $(K-1)/2$ density parameters. On a ResNet classification task the same idea raises test accuracy from 46% (uniform) to 53% (optimised density) while adding about 11% to per-convolution execution time. If these results hold, convolutional networks can be improved without changing architecture simply by replacing the uniform kernel scaling with a small set of learned position weights.

What carries the argument

The central object is the density function $\Phi$, a $K \times K$ matrix that is multiplied element-wise (Hadamard product) with the convolution kernel $w^f$ before the inner product with the image patch, so the weighted convolution reads $(I \ast W_{\Phi})^f_{ij} = \langle \Phi \circ w^f, \mathcal{N}(I_{ij}) \rangle_F$. It is parametrised as $\Phi = \alpha \alpha^{\top}$ with $\alpha$ symmetric about its centre and with the central value fixed, so the number of free density variables is only $(K-1)/2$. The argument is carried by a two-level optimisation: the kernel weights $W$ are learned by stochastic gradient descent, while the density $\alpha$ is learned by the global, derivative-free DIRECT-L method on the same loss function. The separation is the key step, because it lets a cheap local optimiser handle the weights and a derivative-free optimiser handle the non-convex, non-differentiable density functional. The paper also proves that the weighted convolution satisfies a convolution theorem $\mathcal{F}(f \ast g_{\varphi}) = \mathcal{F}(f) \cdot \mathcal{F}(g \cdot \varphi)$, a commutativity relation, a differentiability identity, and Young's inequality, which keeps the operation inside the standard Fourier and measure-theoretic toolbox.

What would settle it

Evaluate the 5×5 weighted convolution with $\alpha = [0.38, 2.21, 1, 2.21, 0.38]$ against uniform convolution on the same denoising network but on a held-out set of noisy images never used to train the density or the kernel; if the 53% loss reduction relative to uniform does not persist on that test set, the claim that the optimal density improves accuracy beyond the training data collapses.

Watch

Extended reading notes

Core claim

The paper's claim is that replacing the standard uniform convolution with a density-weighted convolution — in which each kernel entry is scaled by a symmetric, rank-one density $\Phi = \alpha \alpha^{\top}$ before the Frobenius inner product with the image neighbourhood — improves both the training objective and test accuracy of convolutional networks. The evidence is a three-layer image-denoising network where the optimised densities reduce loss by 12% (3×3), 53% (5×5) and 30% (7×7) relative to uniform, and a ResNet on STL-10 where the optimal density yields 53% test accuracy (on a 30-image test set) versus 46% for uniform. The paper shows that the optimal densities concentrate influence on specific kernel positions (for 5×5, the ring just off-centre is weighted at 2.21 while the corners are weighted at 0.38), and that the density values converge to a stable profile as the model's capacity and dataset grow. The paper also derives analytic properties of the weighted convolution — a convolution theorem, a commutativity identity, differentiability, and Young's integrability bound — showing the operation remains mathematically well-behaved.

Load-bearing premise

The load-bearing premise is that the density function that minimises the loss on the training set also improves performance on new, unseen data; the paper gives no held-out test for the denoising task and only a 30-image test set for classification.

Editorial extensions

If this is right

  • A 5×5 convolution that scales its kernel by $\alpha = [0.38, 2.21, 1, 2.21, 0.38]$ instead of treating all offsets equally lowers the denoising loss by 53% in the paper's experiments, so the effective receptive field matters more than raw kernel size.
  • The optimal density converges to a stable profile as the number of epochs, images, and channels grows, which means the density can be reused after training and does not need to be re-optimised for every hyperparameter setting.
  • Replacing uniform scaling with the optimised density raises test accuracy from 46% to 53% on a ResNet classification task, showing the benefit is not limited to image-to-image regression.
  • Because the weighted convolution only adds one extra multiplication per kernel tap, its execution time is about 11% higher than standard convolution, a small cost that does not change the asymptotic complexity of the network.
  • The analytic properties the paper proves — a convolution theorem, a commutativity identity, differentiability, and Young's integrability bound — mean the weighted convolution can be deployed inside standard CNN architectures without breaking the mathematical foundations that convolution layers rely on.

Reading between the lines

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

  • A natural step the paper does not take is to test whether the optimal density learned on one denoising task transfers to another: if the density's shape reflects the correlation structure of natural images, the same $\Phi$ should work across noise levels and datasets, leaving only the kernel weights to be retrained.
  • Because the density is shared across all kernels of a layer and the optimisation cost grows exponentially in the number of density variables, extending the method to very large kernels would need the rank-one or symmetric structure to be exploited more aggressively, or a cheaper density optimiser.
  • The paper never relaxes the rank-one constraint $\Phi = \alpha \alpha^{\top}$; a full symmetric $\Phi$ would show how much of the gain is due to separability of the density rather than to the density itself, and would be the cleanest next experiment.
  • With only 30 test images in the classification comparison, the reported 46% to 53% accuracy gap is a rough estimate; repeating the experiment on the full 8,000-image STL-10 test set would give a stable measure of the advantage.
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

4 major / 4 minor

Summary. The paper introduces a weighted convolution operator for CNNs in which the convolution kernel is multiplied element-wise by a density matrix Φ=αα^T, with α constrained to be symmetric and to have a fixed central value M. The kernel weights are trained with SGD, while the density coefficients are optimized separately with the global derivative-free solver DIRECT-L on the same training loss. Experiments are reported for an image-denoising architecture with 3×3, 5×5, and 7×7 kernels and for an STL-10 classification task using ResNet. The paper claims up to 53% loss reduction relative to uniform convolution and an increase in classification accuracy from 46% to 53%, together with an analysis of execution time and an appendix on analytic properties of the weighted convolution.

Significance. If the claimed improvements generalized to held-out data, the proposed weighted convolution would be a simple, plug-in modification of standard CNN convolutions with broad applicability, and the public code release would make it easy to reproduce. The analytic properties in the appendix and the computational cost analysis are useful and largely correct. However, the central empirical claim is not supported by the evidence as presented: the main loss reductions are measured on the very training objective being minimized by the density optimizer, and the only test-set evaluation uses 30 images without error bars. The additional capacity brought by the density coefficients is not controlled for, so the reported gains could be an artifact of fitting rather than a genuine generalization improvement. The idea is worth pursuing, but the current manuscript does not validate it.

major comments (4)
  1. [§4.1, Eq. (6)] The reported 12%, 53%, and 30% loss reductions are computed on the same 200-image training objective that DIRECT-L minimizes in Eq. (6). Since the density coefficients α are free variables fitted to that objective, obtaining a lower training loss for the fitted α relative to the uniform α=1 is the expected consequence of the fitting procedure, not evidence of improved denoising performance. No held-out split or test-set loss is reported for the denoising task, so the abstract's claim that the weighted convolution 'significantly reduces the loss' is unsupported for unseen data.
  2. [Table 3] The only test-set evaluation uses 30 STL-10 images. The difference between 53% and 46% classification accuracy corresponds to roughly two images, and no confidence intervals, repeated runs, or per-class breakdowns are given. This result cannot support the claimed test accuracy increase, and the table should either be removed or replaced with a statistically meaningful evaluation.
  3. [§3.3 and §1] The paper repeatedly states that the weighted convolution retains 'the same number of trainable weights.' While the count of kernel weights is unchanged, the density coefficients α are optimized on the training set and therefore add (K−1)/2 free parameters to the model. This is additional capacity that must be controlled in a comparison; the experiments do not compare against a baseline with an equivalent increase in kernel parameters, so the improvement cannot be attributed to the density function rather than to the extra capacity.
  4. [Table 2] The robustness claim that 'the α1 value converges to the optimal value of 0.42' is contradicted by the same table, where α1 ranges from 0.31 to 1.48 across the reported hyperparameter settings. Moreover, each row appears to come from a single optimization run with no repeated seeds or variance measure, so the convergence and robustness claims are not established by the presented data.
minor comments (4)
  1. [§3.2, Eq. (4)] The sentence '⟨A,B⟩_F =: C, C(i,j) := A(i,j)B(i,j)' is confusingly written; the intended relation is simply that the Frobenius inner product is the sum of the entries of the Hadamard product.
  2. [§3.3] The phrase 'given a squared kernel K_a = K_b = K' should read 'given a square kernel' to avoid ambiguity.
  3. [Appendix, Young's inequality] The derivation of Young's inequality contains an informal step after the second inequality where the inner integral is bounded; the application of the triangle inequality and the change of variables should be stated more explicitly.
  4. [§4.1, Fig. 1] The figure caption and the text both state that the 3×3 objective is 'convex', but only a single one-dimensional plot is shown; the paper should avoid a global convexity claim without further evidence.

Circularity Check

2 steps flagged · score 7.0 of 10

The 53% loss reduction is the value of the objective minimized over alpha on the same training set; the only held-out accuracy uses 30 images.

  1. fitted input called prediction [Sect. 3.3, Eq. (6); Sect. 4.1, Fig. 2]
    "The optimal density function [0.38,2.21,1,2.21,0.38] reduces the objective function of the 53% with respect to the uniform density function [1,1,1,1,1]."

    The objective function is the quantity minimized in Eq. (6), min_Phi M_Phi, where M_Phi is the training loss after optimizing W for a fixed Phi. The uniform vector [1,1,1,1,1] is a feasible point of the same optimization variable alpha, so the value at the found density is, by construction, no larger than the value at the uniform density. Reporting the 53% gap as evidence that weighted convolution improves approximation accuracy is reporting the result of the fit, not an independent test; no held-out split is used for the image-to-image denoising task.

  2. self definitional [Sect. 4.1 introductory paragraph; Eq. (6)]
    "The results of our optimisation model show that the optimal density function reaches better results than the uniform density function, with a reduction of the loss function of the learning model for an average value of 30%."

    'Optimal density function' is defined as the minimizer of M_Phi in Eq. (6), and the uniform density is a particular feasible choice of the same variable. The statement that the minimizer attains a lower value than a non-minimizing feasible point is true by definition of optimality and therefore cannot serve as evidence that the weighted convolution itself generalizes better or approximates better outside the training objective. The 12%, 53%, and 30% figures are all comparisons within the fitted objective.

full rationale

The paper's central quantitative claim, 'weighted convolution significantly reduces the loss (up to 53% improvement) and increases the test accuracy', rests on two evidential pillars. The loss pillar is circular: the 53% improvement is the difference between the value of the objective in Eq. (6) at the alpha found by DIRECT-L and the value at the uniform alpha, both evaluated on the same 200-image training set used to select alpha. Since Eq. (6) is defined as minimization over Phi, and the uniform density is a feasible point of that same optimization, the inequality M_Phi(alpha_opt) <= M_Phi(alpha_uniform) holds by construction; it is the outcome of fitting an additional (K-1)/2 parameter, not an independent prediction. The test-accuracy pillar is not circular, but it is statistically thin: Table 3 reports a 53% vs 46% accuracy difference on a test set of 30 STL-10 images, roughly two images, and no held-out evaluation is provided for the denoising task. Self-citations such as [CP25] are present but are not load-bearing for the main result; DIRECT-L is an external algorithm, and the cited work is not used to force the conclusion. Because the headline loss reduction reduces by construction to the optimization objective, the paper warrants a high circularity score, though not the maximum: the density shapes, robustness tables, and small held-out classification comparison contain some independent content.

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

The paper's quantitative results depend on the fitted density coefficients and on arbitrary constraints (symmetric, rank-1, shared density). These constraints are not derived from theory or validated against data. No new entities are introduced.

free parameters (3)
  • alpha_1 (3x3 kernel) = 0.42
    Optimized by DIRECT-L to minimize the training loss; reported in Sec. 4.1.
  • alpha_1, alpha_2 (5x5 kernel) = 0.38, 2.21
    Optimized to minimize training loss; reported in Sec. 4.1.
  • alpha_1, alpha_2, alpha_3 (7x7 kernel) = 0.06, 1.23, 1.72
    Optimized via DIRECT-L; reported in Sec. 4.1.
assumptions (3)
  • ad hoc to paper Density function is symmetric and rank-1 (Phi = alpha alpha^T)
    Reduces the optimization space to (K-1)/2 variables, but is an arbitrary restriction that may bias the result; Sec. 3.3.
  • domain assumption One density function is shared across all filters and spatial locations
    Assumes a global positional preference in the data; may not hold for all tasks; Sec. 3.2.
  • ad hoc to paper DIRECT-L with bounds [0,2M] finds a meaningful optimum
    The global optimizer is run for a limited number of iterations, but the paper does not analyze convergence or optimality; Sec. 3.3.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Optimal Density Functions for Weighted Convolution in Learning Models." pith.science (2026). https://pith.science/paper/KPQ7YFVR

@misc{pith2026250524527,
  author       = {Pith},
  title        = {Pith review of: Optimal Density Functions for Weighted Convolution in Learning Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KPQ7YFVR}},
  note         = {Machine review of arXiv:2505.24527}
}
read the original abstract

The paper introduces the weighted convolution, a novel approach to the convolution for signals defined on regular grids (e.g., 2D images) through the application of an optimal density function to scale the contribution of neighbouring pixels based on their distance from the central pixel. This choice differs from the traditional uniform convolution, which treats all neighbouring pixels equally. Our weighted convolution can be applied to convolutional neural network problems to improve the approximation accuracy. Given a convolutional network, we define a framework to compute the optimal density function through a minimisation model. The framework separates the optimisation of the convolutional kernel weights (using stochastic gradient descent) from the optimisation of the density function (using DIRECT-L). Experimental results on a learning model for an image-to-image task (e.g., image denoising) show that the weighted convolution significantly reduces the loss (up to 53% improvement) and increases the test accuracy compared to standard convolution. While this method increases execution time by 11%, it is robust across several hyperparameters of the learning model. Future work will apply the weighted convolution to real-case 2D and 3D image convolutional learning problems.

Figures

Figures reproduced from arXiv: 2505.24527 by the authors.

Figure 1
Figure 1. Optimisation results on the 3×3 kernel. Left: α1 values (x−axis) with respect to the objective function (y−axis). Centre: α1 optimisation (y−axis) over the iterations of the optimiser (x−axis). Right: objective function (y−axis) over the iterations of the optimiser (x−axis). • Symmetry along both the dimensions, i.e., α = β, α(i) = α(K − i + 1), i = 1 . . . K; • Value of the central node as αm = M. The density funct… view at source ↗
Figure 2
Figure 2. Optimisation results on the 5 × 5 kernel. Left: α1 (x−axis) and α2 (y−axis) values with respect to the objective function: from blue (low) to yellow (high), and black iso-contour of minimal values of the loss function. Centre: α values optimisation (y−axis) over the iterations of the optimiser (x−axis): top graph represents α1, bottom graph represents α2. Right: objective function (y−axis) over the iterations of the… view at source ↗
Figure 3
Figure 3. Optimisation results on the 7 × 7 kernel. Left: α1 (x−axis), α2 (y−axis), and α3 (z−axis) values with respect to the objective function: from blue (low) to yellow (high), and black iso-surface of minimal values of the loss function. Centre: α values optimisation (y−axis) over the iterations of the op￾timiser (x−axis): top graph represents α1, middle graph represents α2, and bottom graph represents α3. Right: objecti… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: 3 × 3 (left), 5 × 5 (centre), and 7 × 7 (right) density function. Nodes on the x−axis, density function values on the y−axis. a larger value on the nodes adjacent to the central one, and the values on the external nodes are lower; the optimal α ∈ R 7 density function h…
Figure 5
Figure 5. Figure 5: (a) Different density functions: uniform (blue); linear with slope equal [PITH_FULL_IMAGE:figures/full_fig_p013_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 26 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

    work page 2022

  2. [2]

    CNN -hyperparameter optimization for diabetic maculopathy diagnosis in optical coherence tomography and fundus retinography

    Ghada Atteia, Nagwan Abdel Samee, El-Sayed M El-Kenawy, and Abdelhameed Ibrahim. CNN -hyperparameter optimization for diabetic maculopathy diagnosis in optical coherence tomography and fundus retinography. Mathematics , 10(18):3274, 2022

    work page 2022

  3. [3]

    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

    2020

  4. [4]

    Discrete weighted transforms and large-integer arithmetic

    Richard Crandall and Barry Fagin. Discrete weighted transforms and large-integer arithmetic. Mathematics of Computation , 62(205):305--324, 1994

    work page 1994

  5. [5]

    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

    2020

  6. [6]

    An analysis of single-layer networks in unsupervised feature learning

    Adam Coates, Andrew Ng, and Honglak Lee. An analysis of single-layer networks in unsupervised feature learning. In Proceedings of the fourteenth international conference on artificial intelligence and statistics , pages 215--223. JMLR Workshop and Conference Proceedings, 2011

    work page 2011

  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

    2022

  8. [8]

    Learning-based low-rank denoising

    Simone Cammarasana and Giuseppe Patane. Learning-based low-rank denoising. Signal, Image and Video Processing , 17(2):535--541, 2023

    work page 2023

Show all 44 references
  1. [9]

    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

    2025

  2. [10]

    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

  3. [11]

    Self-organizing multilayered neural network

    K Fukushima. Self-organizing multilayered neural network. The Transactions of Electronics and Communication Engineers D , 58(9):530, 1975

    1975

  4. [12]

    A locally-biased form of the DIRECT algorithm

    Joerg M Gablonsky and Carl Tim Kelley. A locally-biased form of the DIRECT algorithm. Technical report, North Carolina State University. Center for Research in Scientific Computation, 2000

    2000

  5. [13]

    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

  6. [14]

    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

  7. [15]

    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

  8. [16]

    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

  9. [17]

    Backpropagation and stochastic gradient descent method

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

    1993

  10. [18]

    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

  11. [19]

    A new direction adaptive scheme for image interpolation

    Hao Jiang and Cecilia Moloney. A new direction adaptive scheme for image interpolation. In Proceedings. International Conference on Image Processing , volume 3, pages III--III. IEEE, 2002

    2002

  12. [20]

    Lipschitzian optimization without the L ipschitz constant

    Donald R Jones, Cary D Perttunen, and Bruce E Stuckman. Lipschitzian optimization without the L ipschitz constant. Journal of Optimization Theory and Applications , 79:157--181, 1993

    1993

  13. [21]

    Resnet 50

    Brett Koonce and Brett Koonce. Resnet 50. Convolutional neural networks with swift for tensorflow: image recognition and dataset categorization , pages 63--72, 2021

    2021

  14. [22]

    Deep learning in the automotive industry: Applications and tools

    Andre Luckow, Matthew Cook, Nathan Ashcraft, Edwin Weill, Emil Djerekarov, and Bennie Vorster. Deep learning in the automotive industry: Applications and tools. In International Conference on Big Data , pages 3759--3768. IEEE, 2016

    2016

  15. [23]

    Fast algorithms for convolutional neural networks

    Andrew Lavin and Scott Gray. Fast algorithms for convolutional neural networks. In Proceedings of the IEEE conference on computer vision and pattern recognition , pages 4013--4021, 2016

    2016

  16. [24]

    Weighted adaptive lifting-based wavelet transform for image coding

    Yu Liu and King Ngi Ngan. Weighted adaptive lifting-based wavelet transform for image coding. Transactions on Image Processing , 17(4):500--511, 2008

    2008

  17. [25]

    Omni-dimensional dynamic convolution

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

    2022

  18. [26]

    Convolutional kernel networks

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

    2014

  19. [27]

    Deep learning in robotics: a review of recent research

    Harry A Pierson and Michael S Gashler. Deep learning in robotics: a review of recent research. Advanced Robotics , 31(16):821--835, 2017

    2017

  20. [28]

    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

  21. [29]

    U-net: Convolutional networks for biomedical image segmentation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomedical image segmentation. In Medical image computing and computer-assisted intervention--MICCAI 2015: 18th international conference, Munich, Germany, October 5-9, 2015, proceedings, part...

    2015

  22. [30]

    Ricker, ormsby, klander, butterworth - a choice of wavelets

    Harold Ryan and Hi-Res Geoconsulting. Ricker, ormsby, klander, butterworth - a choice of wavelets. CSEG Recorder , 19(07), 1994

    1994

  23. [31]

    An overview of gradient descent optimization algorithms

    Sebastian Ruder. An overview of gradient descent optimization algorithms. arXiv preprint arXiv:1609.04747 , 2016

    2016 arXiv

  24. [32]

    Deep learning with pytorch: Build, train, and tune neural networks using python tools, 2020

    Eli Stevens, Luca Antiga, and Thomas Viehmann. Deep learning with pytorch: Build, train, and tune neural networks using python tools, 2020

    2020

  25. [33]

    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

  26. [34]

    Very deep convolutional networks for large-scale image recognition

    Karen Simonyan and Andrew Zisserman. Very deep convolutional networks for large-scale image recognition. In 3rd International Conference on Learning Representations (ICLR) , 2015

    2015

  27. [35]

    Mingxing Tan and Quoc V. Le. Efficientnet: Rethinking model scaling for convolutional neural networks, 2020

    2020

  28. [36]

    Catmull-rom splines

    Christopher Twigg. Catmull-rom splines. Computer , 41(6):4--6, 2003

    2003

  29. [37]

    Deep learning for computer vision: A brief review

    Athanasios Voulodimos, Nikolaos Doulamis, Anastasios Doulamis, and Eftychios Protopapadakis. Deep learning for computer vision: A brief review. Computational Intelligence and Neuroscience , 2018(1):7068349, 2018

    2018

  30. [38]

    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

  31. [39]

    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

  32. [40]

    Weighted support vector machine for data classification

    Xulei Yang, Qing Song, and Aize Cao. Weighted support vector machine for data classification. In Proceedings. International Joint Conference on Neural Networks , volume 2, pages 859--864. IEEE, 2005

    2005

  33. [41]

    Kernel-based fuzzy and possibilistic c-means clustering

    Dao-Qiang Zhang and Song-Can Chen. Kernel-based fuzzy and possibilistic c-means clustering. In Proceedings of the international conference artificial neural network , volume 122, pages 122--125, 2003

    2003

  34. [42]

    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

  35. [43]

    A sufficient condition for convergences of adam and rmsprop

    Fangyu Zou, Li Shen, Zequn Jie, Weizhong Zhang, and Wei Liu. A sufficient condition for convergences of adam and rmsprop. In Proceedings of the IEEE/CVF Conference on computer vision and pattern recognition , pages 11127--11135, 2019

    2019

  36. [44]

    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

Pith tools

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