arXiv research

Differential Galois theory connects connections with parameters to isomonodromic deformations.

We assume a vector bundle $p: E\to M$ with a general linear connection $K$ and a classical linear connection $\Lam$ on $M$. We prove that all classical linear connections on the total space $E$ naturally given by $(\Lam, K)$ form a 15-parameter family. Further we prove that all connections on $J^1 E$ naturally given by…

On a main class of the almost contact manifolds with B-metric, it is described the family of the linear connections preserving the manifold's structures by 4 parameters. In this family there are determined the canonical-type connection and the connection with zero parameters.

Kernel for multi-parameter persistent homology connects TDA with ML.

Study improves queue length estimation from connected vehicles by filtering parameters.

We derive reconstruction formulas for a family of geodesic ray transforms with connection, defined on simple Riemannian surfaces. Such formulas provide injectivity of such all transforms in a neighbourhood of constant curvature metrics and non-unitary connections with curvature close to zero. If certain Fredholm equati…

In this note, we extend the theory of Chern-Cheeger-Simons to construct canonical invariants for a one-parameter family of flat connections on a smooth manifold. These invariants lie in degrees $(2p-2)$-cohomology with $\C/\Z$-cohomology, for $p\geq 2$. Furthermore, they are shown to be rigid in a variation of paths (p…

Quantum connections replace metrics with operator inner products.

Stanza separates convolutional and fully connected layers for faster deep learning training.

The study shows removing fully connected output layers improves efficiency without sacrificing performance.

DenseNets improve accuracy and efficiency in convolutional networks.

In this paper we prove the infinitesimal uniqueness theorem for the Newton potential of non simply connected bodies using the singularity theory approach. We consider the Newtonian potentials of the domains in ${\bf R}^n$ boundaries of which are the vanishing cycles on the level hypersurface of a holomorphic function w…

Study explores warped geometries of tensor manifolds, finding non-geodesic connections for some parameters.

We analyze a Lagrangian for spacetime connections in Loop Quantum Gravity.

A new neural network architecture reduces parameters by 94% while maintaining performance.

Law derived for neural networks with sparse connections.

We introduce the equation of n-dimensional totally geodesic submanifolds of a manifold E as a submanifold of the second order jet space of n-dimensional submanifolds of E. Next we study the geometry of n-Grassmannian equivalent connections, that is linear connections without torsion admitting the same equation of n-dim…

Our work connects parameter magnitudes and Hessian eigenspaces in deep neural nets.

Paper constructs solutions to a system using Aeppli class without auxiliary gauge connection.

Empirical study shows removing neural parameter symmetries impacts model performance.

We introduce a new deep convolutional neural network, CrescendoNet, by stacking simple building blocks without residual connections. Each Crescendo block contains independent convolution paths with increased depths. The numbers of convolution layers and parameters are only increased linearly in Crescendo blocks. In exp…

Functional dimension varies in ReLU networks, with implications for symmetry and connectivity.

In four dimensions one can use the chiral part of the spin connection as the main object that encodes geometry. The metric is then recovered algebraically from the curvature of this connection. We address the question of how isometries can be identified in this "pure connection" formalism. We show that isometries are r…

Constructs Lagrangian correspondences for Higgs bundles and holomorphic connections.

Covariant formulation of Barbero-Immirzi connections for spin manifolds.

We consider the problem of multi-task learning in the high dimensional setting. In particular, we introduce an estimator and investigate its statistical and computational properties for the problem of multiple connected linear regressions known as Data Enrichment/Sharing. The between-tasks connections are captured by a…

Constructs moduli stacks for quiver connections and extends non-Abelian Hodge theory.

Mesoscopic model infers neural population dynamics from spike trains.

This paper explains GCNs using NTKs and improves their performance.

Improved traffic flow prediction model using Kalman filter noise reduction.

We systematically discuss connections on the spinor bundle of Cahen-Wallach symmetric spaces. A large class of these connections is closely connected to a quadratic relation on Clifford algebras. This relation in turn is associated to the symmetric linear map that defines the underlying space. We present various soluti…

New connection found between shape reconstruction methods and persistent homology.

AI detects LDDoS attacks by analyzing TCP connection parameters.

We construct many new invariant solutions to the Strominger system with respect to a 2-parameter family of metric connections $\nabla^{\varepsilon,ρ}$ in the anomaly cancellation equation. The ansatz $\nabla^{\varepsilon,ρ}$ is a natural extension of the canonical 1-parameter family of Hermitian connections found by Ga…

Study higher genus polylogarithms under Riemann surface degenerations.

If we consider the moduli space of flat connections of a non trivial principal SO(3)-bundle over a surface, then we can define a map from the set of perturbed closed geodesics, below a given energy level, into families of perturbed Yang-Mills connections depending on a small parameter. In this paper we show that this m…

Structural RBM reduces parameters for image denoising and classification.

Nontrivial connectivity has allowed the training of very deep networks by addressing the problem of vanishing gradients and offering a more efficient method of reusing parameters. In this paper we make a comparison between residual networks, densely-connected networks and highway networks on an image classification tas…

We define the notion of a formal connection for a smooth family of star products with fixed underlying symplectic structure. Such a formal connection allows one to relate star products at different points in the family. This generalizes the formal Hitchin connection introduced by the first author. We establish a necess…

Study on plane curves with special connections and curvatures.

We study several linear connections (the first canonical, the Chern, the well adapted, the Levi Civita, the Kobayashi-Nomizu, the Yano, the Bismut and those with totally skew-symmetric torsion) which can be defined on the four geometric types of $(J^2=\pm1)$-metric manifolds. We characterize when such a connection is a…

Study analyzes error in ReLU networks with local connections.

The fully connected layers of a deep convolutional neural network typically contain over 90% of the network parameters, and consume the majority of the memory required to store the network parameters. Reducing the number of parameters while preserving essentially the same predictive performance is critically important …

Proposes a low-rank deep CNN for multi-task learning.

SWNets optimize DL architectures for faster convergence.

Skip connections improve biologically-inspired learning rules.

The stochastic block model (SBM) is a probabilistic model for community structure in networks. Typically, only the adjacency matrix is used to perform SBM parameter inference. In this paper, we consider circumstances in which nodes have an associated vector of continuous attributes that are also used to learn the node-…

There is a well known one--parameter family of left invariant CR structures on $SU(2)\cong S^3$. We show how purely algebraic methods can be used to explicitly compute the canonical Cartan connections associated to these structures and their curvatures. We also obtain explicit descriptions of tractor bundles and tracto…