Searcharxiv⌕ Search

arXiv subjects

Benjamin Himpel

Publications and source records attributed to Benjamin Himpel.

14 recordsLinked to original sources
arXiv2026

Spectral Convolution on Orbifolds for Geometric Deep Learning

Geometric deep learning (GDL) deals with supervised learning on data domains that go beyond Euclidean structure, such as data with graph or manifold structure. Due to the demand that arises from application-related data, there is a need to identify further topological and geometric structures with which these use cases can be made accessible to machine learning. There are various techniques, such as spectral convolution, that form the basic building blocks for some convolutional neural network-like architectures on non-Euclidean data. In this paper, the concept of spectral convolution on orbifolds is introduced. This provides a building block for making learning on orbifold structured data accessible using GDL. The theory discussed is illustrated using an example from music theory.

cs.LG↗
arXiv2026

Optimizing Tensor Train Decomposition in DNNs for RISC-V Architectures Using Design Space Exploration and Compiler Optimizations

Deep neural networks (DNNs) have become indispensable in many real-life applications like natural language processing, and autonomous systems. However, deploying DNNs on resource-constrained devices, e.g., in RISC-V platforms, remains challenging due to the high computational and memory demands of fully connected (FC) layers, which dominate resource consumption. Low-rank factorization (LRF) offers an effective approach to compressing FC layers, but the vast design space of LRF solutions involves complex trade-offs among FLOPs, memory size, inference time, and accuracy, making the LRF process complex and time-consuming. This paper introduces an end-to-end LRF design space exploration methodology and a specialized design tool for optimizing FC layers on RISC-V processors. Using Tensor Train Decomposition (TTD) offered by TensorFlow T3F library, the proposed work prunes the LRF design space by excluding first, inefficient decomposition shapes and second, solutions with poor inference performance on RISC-V architectures. Compiler optimizations are then applied to enhance custom T3F layer performance, minimizing inference time and boosting computational efficiency. On average, our TT-decomposed layers run 3x faster than IREE and 8x faster than Pluto on the same compressed model. This work provides an efficient solution for deploying DNNs on edge and embedded devices powered by RISC-V architectures.

cs.LG↗
arXiv2022

Geometry of Music Perception

Prevalent neuroscientific theories are combined with acoustic observations from various studies to create a consistent geometric model for music perception in order to rationalize, explain and predict psycho-acoustic phenomena. The space of all chords is shown to be a Whitney stratified space. Each stratum is a Riemannian manifold which naturally yields a geodesic distance across strata. The resulting metric is compatible with voice-leading satisfying the triangle inequality. The geometric model allows for rigorous studies of psychoacoustic quantities like roughness and harmonicity as height functions. In order to show how to use the geometric framework in psychoacoustic studies, concepts for the perception of chord resolutions are introduced and analyzed.

math.DG↗
arXiv2022

Lie groups and Chern-Simons Theory

Witten introduced classical Chern-Simons theory to topology in 1989, when he defined an invariant for knots in 3-manifolds by an integral over a certain infinite-dimensional space, which up to today have not been entirely understood. However, they motivated lots of interesting questions and results in knot theory and low-dimensional topology, as well as the development of entirely new fields. These are lecture notes for a course in Lie groups and Chern-Simons Theory aimed at graduate students.

math.GT↗
arXiv2020

Smoothing Game

We want to introduce another smoothing approach by treating each geometric element as a player in a game: a quest for the best element quality. In other words, each player has the goal of becoming as regular as possible. The set of strategies for each element is given by all translations of its vertices. Ideally, he would like to quantify this regularity using a quality measure which corresponds to the utility function in game theory. Each player is aware of the other players' utility functions as well as their set of strategies, which is analogous to his own utility function and strategies. In the simplest case, the utility functions only depend on the regularity. In more complicated cases this utility function depends on the element size, the curvature, or even the solution to a differential equation. This article is a sketch of a possible game-theoretical approach to mesh smoothing and still on-going research.

cs.GT↗
arXiv2014

The Witten-Reshetikhin-Turaev invariant for links in finite order mapping tori I

We state Asymptotic Expansion and Growth Rate conjectures for the Witten-Reshetikhin-Turaev invariants of arbitrary framed links in 3-manifolds, and we prove these conjectures for the natural links in mapping tori of finite-order automorphisms of marked surfaces. Our approach is based upon geometric quantisation of the moduli space of parabolic bundles on the surface, which we show coincides with the construction of the Witten-Reshetikhin-Turaev invariants using conformal field theory, as was recently completed by Andersen and Ueno.

math.GT↗
arXiv2014

Laplacian smoothing revisited

We show that the driving force behind the regularizing effect of Laplacian smoothing on surface elements is the popular mean ratio quality measure. We use these insights to provide natural generalizations to polygons and polyhedra. The corresponding functions measuring the quality of meshes are easily seen to be convex and can be used for global optimization-based untangling and smoothing. Using a simple backtracking line-search we compare several smoothing methods with respect to the resulting mesh quality. We also discuss their effectiveness in combination with topology modification.

math.OC↗
arXiv2014

Efficient Mesh Optimization Using the Gradient Flow of the Mean Volume

The signed volume function for polyhedra can be generalized to a mean volume function for volume elements by averaging over the triangulations of the underlying polyhedron. If we consider these up to translation and scaling, the resulting quotient space is diffeomorphic to a sphere. The mean volume function restricted to this sphere is a quality measure for volume elements. We show that, the gradient ascent of this map regularizes the building blocks of hybrid meshes consisting of tetrahedra, hexahedra, prisms, pyramids and octahedra, that is, the optimization process converges to regular polyhedra. We show that the (normalized) gradient flow of the mean volume yields a fast and efficient optimization scheme for the finite element method known as the geometric element transformation method (GETMe). Furthermore, we shed some light on the dynamics of this method and the resulting smoothing procedure both theoretically and experimentally.

math.GT↗
arXiv2013

Efficient and Global Optimization-Based Smoothing Methods for Mixed-Volume Meshes

Some methods based on simple regularizing geometric element transformations have heuristically been shown to give runtime efficient and quality effective smoothing algorithms for meshes. We describe the mathematical framework and a systematic approach to global optimization-based versions of such methods for mixed volume meshes. In particular, we identify efficient smoothing algorithms for certain algebraic mesh quality measures. We also provide explicit constructions of potentially useful smoothing algorithms.

cs.CG↗
arXiv2012

The Witten-Reshetikhin-Turaev invariants of finite order mapping tori II

We identify the leading order term of the asymptotic expansion of the Witten-Reshetikhin-Turaev invariants for finite order mapping tori with classical invariants for all simple and simply-connected compact Lie groups. The square root of the Reidemeister torsion is used as a density on the moduli space of flat connections and the leading order term is identified with the integral over this moduli space of this density weighted by a certain phase for each component of the moduli space. We also identify this phase in terms of classical invariants such as Chern-Simons invariants, eta invariants, spectral flow and the rho invariant. As a result, we show agreement with the semiclassical approximation as predicted by the method of stationary phase.

math.GT↗
arXiv2005

A splitting formula for the spectral flow of the odd signature operator on 3-manifolds coupled to a path of SU(2) connections

We establish a splitting formula for the spectral flow of the odd signature operator on a closed 3-manifold M coupled to a path of SU(2) connections, provided M = S cup X, where S is the solid torus. It describes the spectral flow on M in terms of the spectral flow on S, the spectral flow on X (with certain Atiyah-Patodi-Singer boundary conditions), and two correction terms which depend only on the endpoints. Our result improves on other splitting theorems by removing assumptions on the non-resonance level of the odd signature operator or the dimension of the kernel of the tangential operator, and allows progress towards a conjecture by Lisa Jeffrey in her work on Witten's 3-manifold invariants in the context of the asymptotic expansion conjecture.

math.GT↗
arXiv2003

Calderon projector for the Hessian of the perturbed Chern-Simons function on a 3-manifold with boundary

The existence and continuity for the Calderon projector of the perturbed odd signature operator on a 3-manifold is established. As an application we give a new proof of a result of Taubes relating the mod 2 spectral flow of a family of operators on a homology 3-sphere with the difference in local intersection numbers of the character varieties coming from a Heegard decomposition.

math.GT↗