Papers › FlexConv: Continuous Kernel Convolutions with Differentiable Kernel Sizes

FlexConv: Continuous Kernel Convolutions with Differentiable Kernel Sizes

15 Oct 2021ICLR 2022 4arXiv:2110.08059archive 2025-07-28

David W. Romero, Robert-Jan Bruintjes, Jakub M. Tomczak, Erik J. Bekkers, Mark Hoogendoorn, Jan C. van Gemert

When designing Convolutional Neural Networks (CNNs), one must select the size\break of the convolutional kernels before training. Recent works show CNNs benefit from different kernel sizes at different layers, but exploring all possible combinations is unfeasible in practice. A more efficient approach is to learn the kernel size during training. However, existing works that learn the kernel size have a limited bandwidth. These approaches scale kernels by dilation, and thus the detail they can describe is limited. In this work, we propose FlexConv, a novel convolutional operation with which high bandwidth convolutional kernels of learnable kernel size can be learned at a fixed parameter cost. FlexNets model long-term dependencies without the use of pooling, achieve state-of-the-art performance on several sequential datasets, outperform recent works with learned kernel sizes, and are competitive with much deeper ResNets on image benchmark datasets. Additionally, FlexNets can be deployed at higher resolutions than those seen during training. To avoid aliasing, we propose a novel kernel parameterization with which the frequency of the kernels can be analytically controlled. Our novel kernel parameterization shows higher descriptive power and faster convergence speed than existing parameterizations. This leads to important improvements in classification accuracy.

PaperPDFConference PDFCode

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

rjbruin/flexconv officialmentioned in papermentioned on GitHubpytorchNOASSERTION report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

DescriptiveImage ClassificationSequential Image ClassificationTime Series Analysis

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Image Classification CIFAR-10 FlexTCN-7 Percentage correct 92.2 #184 of 265 Archive leaderboard report
Sequential Image Classification Sequential CIFAR-10 FlexTCN-6 Unpermuted Accuracy 80.82% #6 of 13 Archive leaderboard report
Sequential Image Classification Sequential MNIST FlexTCN-4 Permuted Accuracy 98.72% #3 of 30 Archive leaderboard report
Sequential Image Classification Sequential MNIST FlexTCN-6 Unpermuted Accuracy 99.62% #28 of 30 Archive leaderboard report
Sequential Image Classification noise padded CIFAR-10 FlexTCN-6 % Test Accuracy 69.87% #1 of 7 Archive leaderboard report
Time Series Analysis Speech Commands FlexTCN-4 % Test Accuracy 97.73 #3 of 6 Archive leaderboard report
Time Series Analysis Speech Commands FlexTCN-6 % Test Accuracy (Raw Data) 91.73 #6 of 6 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

SPEED

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections