arXiv subjects

Lukas König

Publications and source records attributed to Lukas König.

6 recordsLinked to original sources

arXiv2026

Diffusion in quantum state preparation: From passive cooling to system-bath engineering

We investigate and compare two particle number-conserving protocols for the preparation of a topologically nontrivial state. The first is derived from thermally coupling the system to a cold bath, while the second is based on engineered dissipation. We numerically study the time required to reach the target state as well as its robustness against physically important perturbations. Crucially, in both protocols, the cooling capability is limited by dissipatively induced diffusion processes. The resulting quadratic scaling of the cooling time with system size is also corroborated analytically using mean-field approximations and a purely classical random-walk model. Furthermore, we find that the engineered protocol admits a unique and stable dark state, which contributes to an ongoing discussion regarding the applicability of dissipative state preparation to many-body systems.

cond-mat.quant-gas↗

arXiv2026

LAYUP: Asynchronous decentralized gradient descent with LAYer-wise UPdates

The increasing size of deep learning models has made distributed training across multiple devices essential. Synchronous, centralized methods incur large communication and synchronization overheads. Communication efficient algorithms can reduce these overheads, but often require extra buffers, remain sensitive to stragglers or parameter drift. We present LayUp, an asynchronous decentralized SGD method with layer-wise updates. LayUp asynchronously exchanges incremental layer-wise updates during backpropagation. It uses randomized gossip communication, enabling updates to be applied as soon as they are available without buffering. These design choices reduce parameter drift and improve robustness to stragglers. We establish a theoretical upper bound for the gradient bias introduced by layer-wise updates and prove convergence of LayUp. We empirically validate LayUp on vision and language modeling tasks, showing convergence up to ~32% faster in terms of wall-clock time compared to synchronous data parallel training and up to ~27% faster than comparable communication efficient algorithms while maintaining better task performance. This speed-up is partly due to higher model FLOPs utilization, as we demonstrate. By injecting delays into the communication between workers, we show that LayUp remains robust to stragglers while DDP and other methods degrade in performance. Overall, LayUp provides a novel practical, straggler-robust alternative for distributed training without sacrificing accuracy.

cs.LG↗

arXiv2026

Training event-based neural networks with exact gradients via Differentiable ODE Solving in JAX

Existing frameworks for gradient-based training of spiking neural networks face a trade-off: discrete-time methods using surrogate gradients support arbitrary neuron models but introduce gradient bias and constrain spike-time resolution, while continuous-time methods that compute exact gradients require analytical expressions for spike times and state evolution, restricting them to simple neuron types such as Leaky Integrate and Fire (LIF). We introduce the Eventax framework, which resolves this trade-off by combining differentiable numerical ODE solvers with event-based spike handling. Built in JAX, our frame-work uses Diffrax ODE-solvers to compute gradients that are exact with respect to the forward simulation for any neuron model defined by ODEs . It also provides a simple API where users can specify just the neuron dynamics, spike conditions, and reset rules. Eventax prioritises modelling flexibility, supporting a wide range of neuron models, loss functions, and network architectures, which can be easily extended. We demonstrate Eventax on multiple benchmarks, including Yin-Yang and MNIST, using diverse neuron models such as Leaky Integrate-and-fire (LIF), Quadratic Integrate-and-fire (QIF), Exponential integrate-and-fire (EIF), Izhikevich and Event-based Gated Recurrent Unit (EGRU) with both time-to-first-spike and state-based loss functions, demonstrating its utility for prototyping and testing event-based architectures trained with exact gradients. We also demonstrate the application of this framework for more complex neuron types by implementing a multi-compartment neuron that uses a model of dendritic spikes in human layer 2/3 cortical Pyramidal neurons for computation. Code available at https://github.com/efficient-scalable-machine-learning/eventax.

cs.LG↗

arXiv2024

Agent-Agnostic Centralized Training for Decentralized Multi-Agent Cooperative Driving

Active traffic management with autonomous vehicles offers the potential for reduced congestion and improved traffic flow. However, developing effective algorithms for real-world scenarios requires overcoming challenges related to infinite-horizon traffic flow and partial observability. To address these issues and further decentralize traffic management, we propose an asymmetric actor-critic model that learns decentralized cooperative driving policies for autonomous vehicles using single-agent reinforcement learning. By employing attention neural networks with masking, our approach efficiently manages real-world traffic dynamics and partial observability, eliminating the need for predefined agents or agent-specific experience buffers in multi-agent reinforcement learning. Extensive evaluations across various traffic scenarios demonstrate our method's significant potential in improving traffic flow at critical bottleneck points. Moreover, we address the challenges posed by conservative autonomous vehicle driving behaviors that adhere strictly to traffic rules, showing that our cooperative policy effectively alleviates potential slowdowns without compromising safety.

cs.LG↗

arXiv2023

Comparing Transmission- and Epi-BCARS: A Transnational Round Robin on Solid State Materials

Broadband coherent anti-Stokes Raman scattering (BCARS) is an advanced Raman spectroscopy method that combines the spectral sensitivity of spontaneous Raman scattering (SR) with the increased signal intensity of single-frequency coherent Raman techniques. These two features make BCARS particularly suitable for ultra-fast imaging of heterogeneous samples, as already shown in biomedicine. Recent studies demonstrated that BCARS also shows exceptional spectroscopic capabilities when inspecting crystalline materials like lithium niobate and lithium tantalate, and can be used for fast imaging of ferroelectric domain walls. These results strongly suggest the extension of BCARS towards new imaging applications like mapping defects, strain, or dopant levels, similar to standard SR imaging. Despite these advantages, BCARS suffers from a spurious and chemically unspecific non-resonant background (NRB) that distorts and shifts the Raman peaks. Post-processing numerical algorithms are then used to remove the NRB and to obtain spectra comparable to SR results. Here, we show the reproducibility of BCARS by conducting an internal Round Robin with two different BCARS experimental setups, comparing the results on different crystalline materials of increasing structural complexity: diamond, 6H-SiC, KDP, and KTP. First, we compare the detected and phase-retrieved signals, the setup-specific NRB-removal steps, and the mode assignment. Subsequently, we demonstrate the versatility of BCARS by showcasing how the selection of pump wavelength, pulse width, and detection geometry can be tailored to suit the specific objectives of the experiment. Finally, we compare and optimize measurement parameters for the high-speed, hyperspectral imaging of ferroelectric domain walls in lithium niobate.

cond-mat.mtrl-sci↗

arXiv2023

Building a Knowledge Graph of Distributed Ledger Technologies

Distributed ledger systems have become more prominent and successful in recent years, with a focus on blockchains and cryptocurrency. This has led to various misunderstandings about both the technology itself and its capabilities, as in many cases blockchain and cryptocurrency is used synonymously and other applications are often overlooked. Therefore, as a whole, the view of distributed ledger technology beyond blockchains and cryptocurrencies is very limited. Existing vocabularies and ontologies often focus on single aspects of the technology, or in some cases even just on one product. This potentially leads to other types of distributed ledgers and their possible use cases being neglected. In this paper, we present a knowledge graph and an ontology for distributed ledger technologies, which includes security considerations to model aspects such as threats and vulnerabilities, application domains, as well as relevant standards and regulations. Such a knowledge graph improves the overall understanding of distributed ledgers, reveals their strengths, and supports the work of security personnel, i.e. analysts and system architects. We discuss potential uses and follow semantic web best practices to evaluate and publish the ontology and knowledge graph.

cs.CL↗