arXiv subjects

Joerg H. Mueller

Publications and source records attributed to Joerg H. Mueller.

5 recordsLinked to original sources

arXiv2026

From ex(p) to poly: Gaussian Splatting with Polynomial Kernels

Recent advancements in Gaussian Splatting (3DGS) have introduced various modifications to the original kernel, resulting in significant performance improvements. However, many of these kernel changes are incompatible with existing datasets optimized for the original Gaussian kernel, presenting a challenge for widespread adoption. In this work, we address this challenge by proposing an alternative kernel that maintains compatibility with existing datasets while improving computational efficiency. Specifically, we replace the original exponential kernel with a polynomial approximation combined with a ReLU function. This modification allows for more aggressive culling of Gaussians, leading to enhanced performance across different 3DGS implementations. Our results show a notable performance improvement of 4 to 15% with negligible impact on image quality. We also provide a detailed mathematical analysis of the new kernel and discuss its potential benefits for 3DGS implementations on NPU hardware.

cs.LG↗

arXiv2022

Gradient-based Weight Density Balancing for Robust Dynamic Sparse Training

Training a sparse neural network from scratch requires optimizing connections at the same time as the weights themselves. Typically, the weights are redistributed after a predefined number of weight updates, removing a fraction of the parameters of each layer and inserting them at different locations in the same layers. The density of each layer is determined using heuristics, often purely based on the size of the parameter tensor. While the connections per layer are optimized multiple times during training, the density of each layer remains constant. This leaves great unrealized potential, especially in scenarios with a high sparsity of 90% and more. We propose Global Gradient-based Redistribution, a technique which distributes weights across all layers - adding more weights to the layers that need them most. Our evaluation shows that our approach is less prone to unbalanced weight distribution at initialization than previous work and that it is able to find better performing sparse subnetworks at very high sparsity levels.

cs.LG↗

arXiv2021

DONeRF: Towards Real-Time Rendering of Compact Neural Radiance Fields using Depth Oracle Networks

The recent research explosion around implicit neural representations, such as NeRF, shows that there is immense potential for implicitly storing high-quality scene and lighting information in compact neural networks. However, one major limitation preventing the use of NeRF in real-time rendering applications is the prohibitive computational cost of excessive network evaluations along each view ray, requiring dozens of petaFLOPS. In this work, we bring compact neural representations closer to practical rendering of synthetic content in real-time applications, such as games and virtual reality. We show that the number of samples required for each view ray can be significantly reduced when samples are placed around surfaces in the scene without compromising image quality. To this end, we propose a depth oracle network that predicts ray sample locations for each view ray with a single network evaluation. We show that using a classification network around logarithmically discretized and spherically warped depth values is essential to encode surface locations rather than directly estimating depth. The combination of these techniques leads to DONeRF, our compact dual network design with a depth oracle network as its first step and a locally sampled shading network for ray accumulation. With DONeRF, we reduce the inference costs by up to 48x compared to NeRF when conditioning on available ground truth depth information. Compared to concurrent acceleration methods for raymarching-based neural representations, DONeRF does not require additional memory for explicit caching or acceleration structures, and can render interactively (20 frames per second) on a single GPU.

cs.CV↗

arXiv2016

Optimal Orderings of k-subsets for Star Identification

Finding the optimal ordering of k-subsets with respect to an objective function is known to be an extremely challenging problem. In this paper we introduce a new objective for this task, rooted in the problem of star identification on spacecrafts: subsets of detected spikes are to be generated in an ordering that minimizes time to detection of a valid star constellation. We carry out an extensive analysis of the combinatorial optimization problem, and propose multiple algorithmic solutions, offering different quality-complexity trade-offs. Three main approaches are investigated: exhaustive search (branch and prune), goal-driven (greedy scene elimination, minimally intersecting subsets), and stateless algorithms which implicitly seek to satisfy the problem's goals (pattern shifting, base unrank). In practical terms, these last algorithms are found to provide satisfactory approximations to the ideal performance levels, at small computational costs.

math.OC↗

arXiv2007

Compensation of eddy-current-induced magnetic field transients in a MOT

The design and implementation of a current driver for the quadrupole coil of a magneto-optical trap are presented. The control and the power stages of the driver, both based on a push-pull configuration with high-speed transistors, are separated, allowing for a protection of the control electronics. Moreover, the coil current can be set by an analog voltage, a feature that makes possible both fast and adiabatic switching of the quadrupole magnetic field. In order to compensate magnetic field transients induced by eddy currents, the driver allows a quick reversal of the quadrupole coil current providing in such a way the required magnetic flux compensation. From a fit to the measured magnetic field transients, we have extracted the decay constants which confirm the efficiency of the compensation. Furthermore, we measured the influence of eddy currents on a sample of cold caesium atoms, testing at the same time the ultimate performance of the driver for our applications. The results show that the implemented electronic circuit is able to reduce substantially the transient effects with switching times of the magnetic field below 100 microseconds.

physics.ins-det↗