275003 research outputs found
Sort by
Determining system reliability, taking account of system structure and involvement of previous knowledge Bestimmung der systemzuverlässigkeit unter berücksichtigung der systemstruktur und einbindung von vorwissen
No full text13314
Post-Training Language Models for Continual Relation Extraction
No full textReal-world data, such as news articles, social media posts, and chatbot conversations, is inherently dynamic and non-stationary, presenting significant challenges for constructing real-time structured representations through knowledge graphs (KGs). Relation Extraction (RE), a fundamental component of KG creation, often struggles to adapt to evolving data when traditional models rely on static, outdated datasets. Continual Relation Extraction (CRE) methods tackle this issue by incrementally learning new relations while preserving previously acquired knowledge. This study investigates the application of pre-trained language models (PLMs), specifically large language models (LLMs), to CRE, with a focus on leveraging memory replay to address catastrophic forgetting. We evaluate decoder-only models (eg, Mistral-7B and Llama2-7B) and encoder-decoder models (eg, Flan-T5 Base) on the TACRED and FewRel datasets. Task-incremental fine-tuning of LLMs demonstrates superior performance over earlier approaches using encoder-only models like BERT on TACRED, excelling in seen-task accuracy and overall performance (measured by whole and average accuracy), particularly with the Mistral and Flan-T5 models. Results on FewRel are similarly promising, achieving second place in whole and average accuracy metrics. This work underscores critical factors in knowledge transfer, language model architecture, and KG completeness, advancing CRE with LLMs and memory replay for dynamic, real-time relation extraction
Investigations Of Air Atomized And Coarser Gas-atomized AlSi12 Powders To Evaluate Cost Reduction Potentials For Additive Manufacturing Processes
No full textThe powder is a decisive factor for the final quality of an L-PBF part. Due to constant improvements, productivity continues to rise and more cost efficient processes can be realized. Thereby, the importance of material-related cost-savings increase and become more relevant regarding the overall manufacturing costs. In order to save costs, the processing of uncommon AlSi12 powder qualities was investigated. A broader particle size distribution (20-100 μm) and an irregular shaped powder (20-63 μm, air atomized) were compared with a standard powder (20-63 μm). All powders were characterised and the flowability was determined. In-process spreadability was monitored and determined on a test-rig. The component quality was analysed using density cubes and tensile specimens built up on an EOS M290. To increase the efficiency of the L-PBF process, coarser layer thicknesses were achieved with the coarser powder. For a final evaluation, the cost saving potentials for all powders variants were evaluated
MODEL-BASED DATA AUGMENTATION TO IMPROVE THE PERFORMANCE OF MACHINE-LEARNING DIAGNOSTIC SYSTEMS
No full text545550Machine-learning diagnostic systems are already being used to detect abnormal conditions in electrical equipment. The challenge in machine-learning diagnostic approaches lies in dealing with small training databases of abnormal-condition samples available as the performance of the machine-learning diagnostic systems relies on the quality and quantity of the training data. One possible solution to this problem are data augmentation techniques generating synthetic data from the available data to augment the training data. However, these generic methods do not consider domain-specific knowledge about the diagnosis task. In this paper, a model-based data augmentation approach using computer-implementable, electromechanical models is presented. This approach uses statistical information extracted from the available data to sample model parameters and input variables to generate synthetic normal- and abnormal-condition data. The model-based data augmentation is showcased for detecting a distribution transformer fault. Machine-learning diagnostic systems are trained utilizing the model-based data augmentation and benchmarked with state-of-the-art diagnostic systems. The number of abnormal-condition measurements for training is limited and the performance of all diagnostic systems is analysed. It is shown that the model-based data augmentation is capable of improving the diagnosis accuracy for the analysed diagnostic task especially when only a very small amount of abnormal-condition data is available
How to Get People to Leave: Exploring the Influence of Warning Message Informativity on the Evacuation of Large-Scale Events
No full text361371This paper outlines work in progress on the development of an agent-based model (ABM) of large-scale events employed to investigate the significance of warning message informativeness in the context of evacuations. Unlike other disaster scenarios where the detail within warning messages has been shown to impact response times and decision-making, the effectiveness of such informativeness in large-scale event settings remains underexplored. Our preliminary results indicate that warning message informativity can have a significant effect on decision and evacuation times but other influences on human behaviour such as environmental or social cues also need to be taken into account
One-Pot Incorporation of Quantum Defects into Single-Walled Carbon Nanotubes with Arylazo Sulfonates
No full textSingle-walled carbon nanotubes (SWCNTs) consist of a carbon monolayer and fluoresce in the near-infrared (NIR, 800-2500 nm) region. Introduction of sp3 quantum defects (QDs) creates novel NIR emission features, which promise huge potential for (bio)photonics. However, so far defect chemistry based on diazonium salts is mainly limited to benzene derivatives. The reaction side products also quench SWCNT fluorescence and impair additional surface chemistry. Here, we introduce a photochemical strategy based on arylazo sulfonates to a) create more complex aromatic QDs and b) avoid side products that adsorb and quench. We show that this way coumarin and other QDs are incorporated, which increases overall NIR emission by more than 90% without purification. These materials can be non-covalently modified with biocompatible polymers (polyethyleneglycoles, DNA) without a change in photophysical properties and we demonstrate mechanical (viscosity) and chemical (neurotransmitter dopamine) sensing. This one-pot approach creates QDs in SWCNTs and provides access to advanced hybrid materials for photonic applications.65
CAN CONFORMAL PREDICTION OBTAIN MEANINGFUL SAFETY GUARANTEES FOR ML MODELS?
No full textConformal Prediction (CP) has been recently proposed as a methodology to calibrate the predictions of Machine Learning (ML) models so that they can output rigorous quantification of their uncertainties. For example, one can calibrate the predictions of an ML model into prediction sets, that guarantee to cover the ground truth class with a probability larger than a specified threshold. In this paper, we study whether CP can provide strong statistical guarantees that would be required in safety-critical applications. Our evaluation on the ImageNet demonstrates that using CP over state-of-the-art models fails to deliver the required guarantees. We corroborate our results by deriving a simple connection between the CP prediction sets and top-k accuracy
50Prozent - Monolithische III-V-Mehrfachsolarzellen mit über 50% Wirkungsgrad unter konzentrierter Einstrahlung
No full textTowards Integrated 3D Microbatteries: Study of LiPON on Porous Electrodes
No full text885888The demand for higher surface and energy densities in 3D all-solid-state microbatteries (ASSMB) is increasing. The state-of-the-art primarily involves thin film batteries, in which the 3D aspect is achieved by deposition of solid-state electrodes and electrolytes on high aspect ratio structures, such as nanotubes or finger-like structures. Our aim is to develop 3D powder-based electrodes for substrate-integrated microbatteries by mimicking the electrode fabrication in macrobatteries. Here, lithium phosphorous oxynitride (LiPON) is used as a solid-state electrolyte on porous powder-based electrodes. The study demonstrates the successful solidification of electrode powders by LiPON thin films
Approach for the numerical simulation of the machining behavior of WC-Co cemented carbide during grinding
No full text596601Tungsten carbide cobalt (WC-Co) cemented carbides are two-phase, brittle-hard materials whose properties can be adapted to the application via the chemical composition. Due to the high requirements on form and dimensional tolerances of cemented carbide products, the material is mostly machined by grinding. In order to analyze the machining behavior when grinding cemented carbide, a costly and time-consuming analogy test is carried out, the so-called single-grain scratching. To substitute the time-consuming experiments, this paper focuses on the development of an approach for the numerical simulation of single-grain scratching of cemented carbide