Abstract:Convolutional neural networks (CNNs) have become an essential tool for solving many machine vision and machine learning problems. A major element of these networks is the convolution operator which essentially computes the inner product between a weight vector and the vectorized image patches extracted by sliding a window in the image planes of the previous layer. In this paper, we propose two classes of surrogate functions for the inner product operation inherent in the convolution operator and so attain two generalizations of the convolution operator. The first one is based on the class of positive definite kernel functions where their application is justified by the kernel trick. The second one is based on the class of similarity measures defined according to some distance function. We justify this by tracing back to the basic idea behind the neocognitron which is the ancestor of CNNs. Both of these methods are then further generalized by allowing a monotonically increasing function (possibly depending on the weight vector) to be applied subsequently. Like any trainable parameter in a neural network, the template pattern and the parameters of the kernel/distance function are trained with the back-propagation algorithm. As an aside, we use the proposed framework to justify the use of sine activation function in CNNs. Additionally, we discovered a family of generalized convolution operators which is based on the convex combination of the dot-product and the negative squared Euclidean distance functions. Our experiments on the MNIST dataset show that the performance of ordinary CNNs can be achieved by generalized CNNs based on weighted L1/L2 distances, proving the applicability of the proposed generalization of the convolutional neural networks.

A Geometric Framework for Convolutional Neural Networks

Understanding Geometry of Encoder-Decoder CNNs

A Unified Framework for Convolution-based Graph Neural Networks

A Geometric Approach of Gradient Descent Algorithms in Linear Neural Networks

Geometric operations implemented by conformal geometric algebra neural nodes

Geometric Operator Convolutional Neural Network

On the Geometry and Optimization of Polynomial Convolutional Networks

GeometricImageNet: Extending convolutional neural networks to vector and tensor images

Geometric Deep Learning: Grids, Groups, Graphs, Geodesics, and Gauges

How deep learning works --The geometry of deep learning

Breaking the Curse of Dimensionality with Convex Neural Networks

A Unified Framework for Training Neural Networks

Geom-GCN: Geometric Graph Convolutional Networks

On the Geometry of Deep Learning

Understanding over-parameterized deep networks by geometrization

Decoupled Networks ( with Appendix )

Generalizing the Convolution Operator in Convolutional Neural Networks

Parameter Convex Neural Networks

Randomized Geometric Algebra Methods for Convex Neural Networks

A singular Riemannian Geometry Approach to Deep Neural Networks III. Piecewise Differentiable Layers and Random Walks on $n$-dimensional Classes