Paper proposes multiscale self-attentive convolutions for vision and language.
problem Improving language and vision understanding models using self-attention.
method Developed 1D and 2D Self Attentive Convolutions (SAC), multiscale SAC (MSAC).
result MSAC enhances model performance for vision and language tasks.
Graph convolutional networks adapt the architecture of convolutional neural networks to learn rich representations of data supported on arbitrary graphs by replacing the convolution operations of convolutional neural networks with graph-dependent linear operations. However, these graph-dependent linear operations are d…
This paper completes the construction of arbitrary order conformally invariant differential operators in higher spin spaces. Jan Slovák has classified all conformally invariant differential operators on locally conformally flat manifolds. We complete his results in higher spin theory by giving explicit expressions for …
Study learns convolution operators on compact Abelian groups using regularization.
problem Learning convolution operators on compact Abelian groups.
method Regularization-based approach with ridge regression estimator.
result Characterizes the accuracy of the estimator in terms of finite sample bounds.
Convolution and pooling improve kernel methods in image classification.
problem Understanding the interplay between approximation and generalization in convolutional architectures.
method Characterized RKHS of kernels with convolution, pooling, and downsampling, computed generalization error.
result Convolution and pooling operations trade off approximation with generalization power.
Convolution has been playing a prominent role in various applications in science and engineering for many years. It is the most important operation in convolutional neural networks. There has been a recent growth of interests of research in generalizing convolutions on curved domains such as manifolds and graphs. Howev…
We review the properties of transversality of distributions with respect to submersions. This allows us to construct a convolution product for a large class of distributions on Lie groupoids. We get a unital involutive algebra $\cE\_{r,s}'(G,Ω^{1/2})$ enlarging the convolution algebra C∞_c(G,Ω1/2) associate…
LeanResNet reduces CNN complexity without sacrificing accuracy.
problem High computational cost in CNNs, especially in ResNets.
method Introduces lean convolution operators that reduce parameters and complexity.
result LeanResNet achieves comparable results to other reduced architectures with fewer parameters.
Inner product-based convolution has been a central component of convolutional neural networks (CNNs) and the key to learning visual representations. Inspired by the observation that CNN-learned features are naturally decoupled with the norm of features corresponding to the intra-class variation and the angle correspond…
BiGraphNet generalizes graph neural networks for more efficient operations.
problem Fragmented graph neural network architectures hinder optimization.
method Explicitly separates input and output nodes, enabling new efficient operations.
result BiGraphNet accelerates and scales computations in hierarchical networks.
LeanConvNets reduces CNN costs without sacrificing accuracy.
problem High computational cost in fully-coupled CNNs.
method Sparsifying fully-coupled operators in CNNs.
result LeanConvNets achieve similar accuracy to state-of-the-art networks but with lower computational cost.
Defines a new algebra for singular foliations, extending Schwartz kernels.
problem Extending Schwartz kernel operators to singular foliations.
method Defines convolution algebra of transverse distributions, proves representation as operators on spaces of functions.
result Generalizes Schwartz kernel operators to singular foliations.
Convolution Neural Network (CNN) has gained tremendous success in computer vision tasks with its outstanding ability to capture the local latent features. Recently, there has been an increasing interest in extending convolution operations to the non-Euclidean geometry. Although various types of convolution operations h…
Proposes a new convolutional neural network for non-grid data.
problem Limited applicability of standard CNNs to non-grid structured data.
method Introduces Parametric Continuous Convolution (PCC) with learnable kernel functions.
result Significant improvement in point cloud segmentation and lidar motion estimation.
Paper studies a universal logical operator for deep networks, improving interpretability.
problem Learning a universal logical operator for deep convolution networks without manual prescription.
method Exploration of different logical operators (AND, OR, XOR) and learning a universal one.
result Insightful observations lead to a novel logical interpretation of deep convolution networks.
New hypergraph operators improve graph neural networks for higher-order relationships.
problem Learning deep embeddings on high-order graph-structured data.
method Introducing hypergraph convolution and hypergraph attention operators.
result Extensive experimental results show the effectiveness of hypergraph operators.
Paper presents efficient algorithms for convolutional neural networks using Winograd minimal filtering.
problem Resource-efficient implementation of convolutional neural networks.
method Winograd minimal filtering trick applied to M-tap filters (M=3,5,7,9,11) for parallel hardware implementation.
result Approximately 30% reduction in multipliers for fully parallel hardware implementation.
CAOL learns filters from large datasets, and this paper analyzes its performance.
problem Understanding how dataset size impacts CAOL filter updates.
method Analyzes the impact of dataset size on CAOL filter updates using deterministic and probabilistic bounds.
result Empirical evidence suggests that using more training data can improve CAOL performance.
A new linear GCN model improves recommendation performance for large graphs.
problem Training difficulties and over-smoothing in GCN-based CF models.
method Proposes a linear residual graph convolutional network (LRGCCF) to address training difficulties and over-smoothing issues.
result The proposed model yields better efficiency and effectiveness on real datasets.
Introduces Finslerian convolution metrics and their properties.
problem No specific problem stated; focuses on new metric concept.
method Definition and study of Finslerian convolution metrics.
result Characterization of Finslerian convolution metrics of Riemannian, Minkowskian, and Randers types.
GCNs improve regression tasks by aggregating neighbor signals.
problem GCNs' statistical properties in regression tasks are poorly understood.
method Examined two GCN convolutions and their impact on learning error.
result GCNs have a bias-variance trade-off that depends on neighborhood size and topology.
Novel volumetric convolution for unit ball improves 3D object recognition.
problem Efficiently convolving functions in a unit ball for deep learning.
method Developed volumetric convolution using Zernike polynomials.
result Improved 3D object recognition through novel convolution.
New bounds for quantile aggregation unify and clarify existing methods.
problem Analytical bounds for quantile aggregation with dependence uncertainty.
method Using inf-convolution of quantile-based risk measures, establish new analytical bounds called convolution bounds.
result Convolution bounds are the best available and provide sharp results in many cases.
Polygonal meshes provide an efficient representation for 3D shapes. They explicitly capture both shape surface and topology, and leverage non-uniformity to represent large flat regions as well as sharp, intricate features. This non-uniformity and irregularity, however, inhibits mesh analysis efforts using neural networ…
Convolutional operator learning is gaining attention in many signal processing and computer vision applications. Learning kernels has mostly relied on so-called patch-domain approaches that extract and store many overlapping patches across training signals. Due to memory demands, patch-domain methods have limitations w…
New multigrid approach reduces CNN parameters by focusing on structured convolutions.
problem Redundancy in standard CNNs leads to high parameter count.
method Replace standard convolutions with structured multilevel convolutions.
result Linearly proportional number of parameters to network width, no loss in accuracy.
Unified framework for convolution and attention models.
problem Complex neural network structures and their parameter control.
method Unified framework for convolution and attention models.
result Attention models are a special case of convolution with adaptive structure.
Initialization of parameters in deep neural networks has been shown to have a big impact on the performance of the networks (Mishkin & Matas, 2015). The initialization scheme devised by He et al, allowed convolution activations to carry a constrained mean which allowed deep networks to be trained effectively (He et al.…
The paper extends Laplacian spectra approximations to vector bundles.
problem Approximating the spectrum of the connection Laplacian.
method Extending the graph connection Laplacian to vector bundles and proving spectrum approximation.
result The spectrum of the extended operator approximates the spectrum of the connection Laplacian.
Rotation intertwining maps from the set of convex bodies in Rn into itself that are continuous linear operators with respect to Minkowski and Blaschke addition are investigated. The main focus is on Blaschke-Minkowski homomorphisms. We show that such maps are represented by a spherical convolution operator. An applicat…
SWAG uses graph convolutions to recommend products.
problem Product recommendation using graph-structured data.
method Graph Convolutional Network (GCN) with weighted random walks and aggregations.
result Graph-based approach improves product recommendations.
Improved generative models using flexible convolutions.
problem Generating high-quality images efficiently.
method Generalized 1 x 1 convolutions to d x d convolutions, chaining autoregressive and periodic convolutions.
result Flexible d x d convolutions significantly improve generative flow models' performance.
A new method for comparing image probability measures using convolution operators.
problem Efficiently comparing images using conventional sliced Wasserstein methods.
method Proposed convolution sliced Wasserstein (CSW) methods with stride, dilation, and non-linear activation.
result CSW demonstrates favorable performance over conventional sliced Wasserstein in image comparison and deep generative modeling.
New framework for manifold convolutions using toric embeddings.
problem Computational intractability of manifold convolutions.
method Isometric embeddings into tori for global manifold convolutions.
result Global definition of manifold convolutions on finite approximations.
Designs CNNs for better image reconstruction.
problem Image reconstruction from limited data.
method Parseval convolution operators and chaining of elementary modules.
result CNN-based algorithm yields better results than sparsity-based methods.
Proposes continuous convolution layers for flexible feature map resizing.
problem Fixed stride limitations in discrete convolution layers.
method Introduces Continuous Convolution (CC) layers that use learned continuous functions.
result Dynamic and consistent resizing of feature maps at any scale, non-integer and axis-dependent.
HKConv learns hyperbolic features by aggregating kernel points.
problem Challenges in learning good hyperbolic representations using Euclidean operations.
method Proposes HKConv, a trainable hyperbolic convolution that correlates local features with kernel points and aggregates them.
result HKConv learns expressive local features according to hyperbolic geometry and enjoys equivariance to permutation and invariance to parallel transport.
Local convolutions bias neural networks towards high-frequency adversarial examples.
problem High-frequency adversarial examples in neural networks.
method Analysis of different linear and nonlinear architectures, focusing on the impact of local convolution operations.
result Local convolutions induce an implicit bias towards high frequency features, leading to high-frequency adversarial examples.
Network deconvolution removes redundant data to improve neural network performance.
problem Redundant data in neural networks makes training challenging.
method Network deconvolution optimally removes pixel-wise and channel-wise correlations before each layer.
result Network deconvolution improves performance across various datasets.
New 3D protein analysis methods improve accuracy.
problem Lack of suitable learning algorithms for protein data.
method Intrinsic-Extrinsic Convolution and Pooling for 3D protein structures.
result Outperforms state-of-the-art methods on protein analysis tasks.
We characterize the singular values of the linear transformation associated with a standard 2D multi-channel convolutional layer, enabling their efficient computation. This characterization also leads to an algorithm for projecting a convolutional layer onto an operator-norm ball. We show that this is an effective regu…
We present a new convolutional neural network-based time-series model. Typical convolutional neural network (CNN) architectures rely on the use of max-pooling operators in between layers, which leads to reduced resolution at the top layers. Instead, in this work we consider a fully convolutional network (FCN) architect…
Paper revisits graph-CNNs using Laplace-Beltrami spectral filters and polynomials.
problem Improving spectral graph convolutional neural networks (graph-CNNs).
method Developed Laplace-Beltrami CNN (LB-CNN) by replacing graph Laplacian with LB operator and approximating spectral filters using Chebyshev, Laguerre, and Hermite polynomials.
result Classification accuracy of LB-CNN is not dependent on the type of polynomials or operators.
Proposes a method to constrain singular values of convolutional kernels in neural networks.
problem Avoiding exploding/vanishing gradient problems and improving generalizability in neural networks.
method Introduces a penalty function to constrain singular values of convolutional kernels around 1, and derives an algorithm for optimization.
result Demonstrates the effectiveness of the method through numerical examples.
Automates neural network design for diverse tasks.
problem Designing neural networks for new, under-explored domains.
method Introduces XD-Operations and a weight-sharing scheme to transform standard backbones into search spaces of operations.
result Models using XD-Operations achieve lower error than baseline and expert-designed networks on diverse tasks.
BankGCN improves graph convolution networks by handling multi-channel signals with adaptive filter banks.
problem Handling multi-channel graph signals with limited architectures.
method BankGCN decomposes multi-channel signals into subspaces and uses adapted filters for each subspace.
result BankGCN achieves excellent performance in graph classification on benchmark datasets.
New bounds for CNNs show better generalization than previous models.
problem Improving understanding of CNNs' generalization ability.
method Proposed tighter generalization bounds for CNNs by exploiting the sparse and permutation structure of weight matrices and spectral norms of convolution operations.
result Theoretical and experimental results show tighter bounds for CNNs than existing bounds.
Functor connects Lie groupoid algebras to bornological structures.
problem Establishing a functorial relationship between Lie groupoid convolution algebras and bornological structures.
method Developed a monoidal functor from differentiable stacks to Morita 2-category of complete bornological algebras.
result Convolution algebras are self-induced and convolution modules are smooth.