arXiv subjects

Magnus Bengtsson

Publications and source records attributed to Magnus Bengtsson.

5 recordsLinked to original sources

arXiv2026

Continuous Temporal Representations of Event-Based Signals via Interference-Based Wave Modeling

Spatio-temporal signals arising from event-driven biological processes, such as surface electromyography (sEMG), exhibit asynchronous and highly structured activation patterns that are challenging to model using conventional discrete or purely real-valued representations. In this work, we propose a continuous temporal modeling framework based on interference-based wave representations. The approach maps event-like input signals into a complex-valued latent wave field, where temporal structure is encoded through phase modulation and interactions between latent components. By projecting the resulting wave field onto an energy domain, the model induces structured activation patterns that capture both temporal localization and relational dependencies within finite observation windows, without relying on explicit recurrence or causal state propagation. The proposed formulation is particularly suited for event-driven biosignals, where continuous representations enable efficient gradient-based optimization and robust feature extraction. In particular, the method is designed to support learning from sEMG data for downstream control tasks in biomechanical systems, such as prosthetic devices and exoskeletons. Experimental results demonstrate that the proposed interference-based wave model provides improved representation quality compared to purely real-valued representations, while maintaining computational efficiency suitable for practical deployment.

cs.LG↗

arXiv2025

Compressing Large Language Models with PCA Without Performance Loss

We demonstrate that Principal Component Analysis (PCA), when applied in a structured manner, either to polar-transformed images or segment-wise to token sequences, enables extreme compression of neural models without sacrificing performance. Across three case studies, we show that a one-layer classifier trained on PCA-compressed polar MNIST achieves over 98 percent accuracy using only 840 parameters. A two-layer transformer trained on 70-dimensional PCA-reduced MiniLM embeddings reaches 76.62 percent accuracy on the 20 Newsgroups dataset with just 81000 parameters. A decoder-only transformer generates coherent token sequences from 70-dimensional PCA embeddings while preserving over 97 percent cosine similarity with full MiniLM representations, using less than 17 percent of the parameter count of GPT-2. These results highlight PCA-based input compression as a general and effective strategy for aligning model capacity with information content, enabling lightweight architectures across multiple modalities.

cs.CE↗

arXiv2025

C2G-KD: PCA-Constrained Generator for Data-Free Knowledge Distillation

We introduce C2G-KD, a data-free knowledge distillation framework where a class-conditional generator is trained to produce synthetic samples guided by a frozen teacher model and geometric constraints derived from PCA. The generator never observes real training data but instead learns to activate the teacher's output through a combination of semantic and structural losses. By constraining generated samples to lie within class-specific PCA subspaces estimated from as few as two real examples per class, we preserve topological consistency and diversity. Experiments on MNIST show that even minimal class structure is sufficient to bootstrap useful synthetic training pipelines.

cs.LG↗

arXiv2025

Warehouse storage and retrieval optimization via clustering, dynamical systems modeling, and GPU-accelerated routing

This paper introduces a warehouse optimization procedure aimed at enhancing the efficiency of product storage and retrieval. By representing product locations and order flows within a time-evolving graph structure, we employ unsupervised clustering to define and refine compact order regions, effectively reducing picking distances. We describe the procedure using a dynamic mathematical model formulated using tools from random dynamical systems theory, enabling a principled analysis of the system's behavior over time even under random operational variations. For routing within this framework, we implement a parallelized Bellman-Ford algorithm, utilizing GPU acceleration to evaluate path segments efficiently. To address scalability challenges inherent in large routing graphs, we introduce a segmentation strategy that preserves performance while maintaining tractable memory requirements. Our results demonstrate significant improvements in both operational efficiency and computational feasibility for large-scale warehouse environments.

math.OC↗

arXiv2021

A proposed method using GPU based SDO to optimize retail warehouses

Research in warehouse optimization has gotten increased attention in the last few years due to e-commerce. The warehouse contains a waste range of different products. Due to the nature of the individual order, it is challenging to plan the picking list to optimize the material flow in the process. There are also challenges in minimizing costs and increasing production capacity, and this complexity can be defined as a multidisciplinary optimization problem with an IDF nature. In recent years the use of parallel computing using GPGPUs has become increasingly popular due to the introduction of CUDA C and accompanying applications in, e.g., Python. In the case study at the company in the field of retail, a case study including a system design optimization (SDO) resulted in an increase in throughput with well over 20% just by clustering different categories and suggesting in which sequence the orders should be picked during a given time frame. The options provided by implementing a distributed high-performance computing network based on GPUs for subsystem optimization have shown to be fruitful in developing a functioning SDO for warehouse optimization. The toolchain can be used for designing new warehouses or evaluating and tuning existing ones.

cs.DC↗