Fraunhofer-Institut für System- und Innovationsforschung ISI

Filters

1
Reset filters

Settings
Now showing 1 - 10 of 915
  • Publication
    In-Memory SAT-Solver for Self-Verification of Programmable Memristive Architectures
    ( 2024)
    Shirinzadeh, Fatemeh
    ;
    Deb, Arighna
    ;
    Shirinzadeh, Saeideh orcid-logo
    ;
    Kole, Abhoy
    ;
    Datta, Kamalika
    ;
    Drechsler, Rolf
    Formal verification of programmable memristive architectures utilizing emerging nonvolatile memory technologies such as Resistive Random-Access Memory (RRAM) has only been recently addressed by a few works at the software level. In this paper we propose an in-memory SAT solver utilizing inherent analog features of RRAM that enables formal verification of arbitrary designs within resistive crossbars. More importantly, this allows self-verification of in-memory implementations as the correctness of designs can be dynamically checked. Additionally, the required architecture is presented, along with a complexity analysis for latency and hardware overheads
  • Publication
    HyperPIE: Hyperparameter Information Extraction from Scientific Publications
    ( 2024)
    Saier, Tarek
    ;
    Ohta, Mayumi
    ;
    Asakura, Takuto
    ;
    Färber, Michael
    Automatic extraction of information from publications is key to making scientific knowledge machine-readable at a large scale. The extracted information can, for example, facilitate academic search, decision making, and knowledge graph construction. An important type of information not covered by existing approaches is hyperparameters. In this paper, we formalize and tackle hyperparameter information extraction (HyperPIE) as an entity recognition and relation extraction task. We create a labeled data set covering publications from a variety of computer science disciplines. Using this data set, we train and evaluate BERT-based fine-tuned models as well as five large language models: GPT-3.5, GALACTICA, Falcon, Vicuna, and WizardLM. For fine-tuned models, we develop a relation extraction approach that achieves an improvement of 29% F1 over a state-of-the-art baseline. For large language models, we develop an approach leveraging YAML output for structured data extraction, which achieves an average improvement of 5.5% F1. in entity recognition over using JSON. With our best performing model we extract hyperparameter information from a large number of unannotated papers, and analyze patterns across disciplines.
  • Publication
    VADIS - A Variable Detection, Interlinking and Summarization System
    ( 2024)
    Kartal, Yavuz Selim
    ;
    Ahsan Shahid, Muhammad
    ;
    Takeshita, Sotaro
    ;
    Tsereteli, Tornike
    ;
    Zielinski, Andrea
    ;
    Zapilko, Benjamin
    ;
    Mayr, Philipp
    The VADIS system addresses the demand of providing enhanced information access in the domain of the social sciences. This is achieved by allowing users to search and use survey variables in context of their underlying research data and scholarly publications which have been interlinked with each other.
  • Publication
    Fair, Reliable, Independent and Cost-Effective? Preferences of German citizens regarding the design of the energy transition
    ( 2023)
    Burghard, Uta orcid-logo
    ;
    Breitschopf, Barbara orcid-logo
    The German energy transition (ET) comprises a bundle of activities such as increase in energy efficiency, expansion of renewable energy supplies, nuclear and fossil fuel phase[1]out, grid extension and enforcement, roll-out of smart grid and energy storage. Studies show a high general approval of the energy transition and its goals in Germany. However, what does the German population think about the different activities of the energy transition? And can groups of individuals be formed on the basis of the perceptions of these various activities? What is the socio-demographic structure of these groups? We aim to answer these questions based on a survey study in Germany (n=889) with a focus on the perception of so-called design elements of the energy transition, i.e. overall objectives and activities to achieve the energy transition. The data analysis shows a high cost sensitivity in society with regard to energy and a positive evaluation of energy independence. Measures, i.e. different types of policy instruments promoting the energy transition, achieved the least approval. A segmentation based on the perceptions of the design elements revealed four clusters: (1) "rejectors of the energy transition", (2) "energy transition enthusiasts", (3) "reserved environmental promoters" and (4) "price-sensitive supporters of energy independence". The clusters indicated that individuals struggle to comprehend and assess each specific design element independently; instead, they tend to hold a broad positive or negative attitude and evaluation of these elements.
  • Publication
    Verification of In-Memory Logic Design using ReRAM Crossbars
    ( 2023)
    Datta, Kamalika
    ;
    Deb, Arighna
    ;
    Shirinzadeh, Fatemeh
    ;
    Kole, Abhoy
    ;
    Shirinzadeh, Saeideh orcid-logo
    ;
    Drechsler, Rolf
    Resistive Random Access Memory (ReRAM) technologies enable the development of innovative architectures for in-memory computing. Many logic design styles, like Imply, Magic or Majority, have been explored for mapping Boolean functions to ReRAM crossbars. However, little attention has been given to the verification of the mapping process. Simulation based approaches can be used to check the functional correctness of smaller designs, but only formal verification techniques can ensure completeness for larger designs. Some initial works in this area have been proposed, which specifically focus on the verification of micro-operations using majority-based logic design. However, these techniques cannot be directly applied to other logic design styles, like Imply or Magic. This necessitates the design and exploration of more general verification techniques for logic-in-memory using ReRAM crossbars, and opens up the scope for further investigation. In this paper, we provide an overview of existing verification techniques for logic-in-memory designs, and also discuss directions for future work.
  • Publication
    Assessing AI-readiness in production - A conceptual approach
    ( 2023)
    Heimberger, Heidi orcid-logo
    ;
    Horvat, Djerdj
    ;
    Schultmann, Frank
    Due to its high potential to perform many tasks faster, more accurately and in greater detail than humans artificial intelligence (AI) has been attracting growing attention across industries. In manufacturing, AI, in combination with digital sensors, networks and software-based automation, defines a new industrialization age. The integration of AI into production processes promises to boost the productivity, efficiency, as well as the automation of processes. However, AI adoption in manufacturing is currently still in its early stage and lacks practical experiences. This raises the question, to which extent manufacturing companies are ready to implement AI. While approaches to assessing the maturity in terms of the digitalization or Industry 4.0 (I4.0) of manufacturing companies are well established and discussed in the literature, approaches that specifically address AI in manufacturing are still lacking. To address this gap, we present an approach to analyze and monitor the readiness of manufacturing firms for working with AI technologies. In accordance with the existing assessment concepts of digitalization and I4.0, our approach examines different areas of digital technologies on the product and production level of manufacturing firms. Moreover, it incorporates the key foundation for AI-security and data - into a conceptual model. We generally assume that companies need to achieve a certain level of digital readiness in three key dimensions in order to be ready for implementing AI-based technologies. We operationalize these dimensions through a variety of product- and production-specific as well as data- and safety-related indicators. In order to illustrate the implementation of our concept in practical terms, we present the results of the readiness assessment of two German manufacturing companies.
  • Publication
    Self-consumption rises due to energy crises? An evaluation of prosumers’ consumption behavior in 2022
    ( 2023)
    Pelka, Sabine orcid-logo
    ;
    Conradie, Peter
    ;
    Vries, Laurens de
    ;
    Anatolitis, Vasilios orcid-logo
    ;
    Martens, Emma
    ;
    Chappin, Emile
    ;
    Karaliopoulos, Merkouris
    ;
    Anagnostopoulos, Filippos
    ;
    Preuß, Sabine orcid-logo
    Prosumers with photovoltaic systems can reduce their electricity expenses by increasing their consumption of self-generated electricity. This makes them more resilient to price shocks, like the 2022 European energy crisis. We evaluate how prosumers adapt their consumption behavior in response to such political uncertainty and increasing electricity prices. The collected survey and smart meter data allow us to evaluate the perceived self-reported and measured impact on self-consumption. Saving intentions due to the energy crisis are more clearly displayed by the survey than by the measured self-consumption. While solar radiation predominantly explains self-consumption changes, Google searches on electricity-related topics have limited explanatory power. However, considering time lags and the interaction with solar radiation leads to more nuanced insights on the effect of Google searches. Depending on the level of solar radiation, the effect of Google searches ranges from decreasing the daily self-consumption by 26.45 Wh to increasing it by 69.45 Wh.
  • Publication
    How to Assess the Impact of Blockchain on Decarbonization in Urban Logistics?
    ( 2023)
    Samet, Mehdi Jahangir
    ;
    Köhler, Jonathan orcid-logo
    ;
    Horák, T.
    The critical role of blockchain technology has been highlighted in decarbonization and logistics transitions in the literature. Blockchain technology combined with the Smart City paradigm is identified as one of the most important digital technology disruptions and trends of sustainable urban logistics in the future. Unfortunately, none of the current strategic assessment models can evaluate the impact of blockchain technology adoption on the decarbonization pathways in urban logistics. In this study, to assess the impact of blockchain technology adoption on decarbonization goals in urban logistics, we review the literature for the current strategic assessment tools/models, sustainable urban logistics, Smart City paradigm, and blockchain technology application in logistics and decarbonization. We propose a combination of different modelling approaches including the living lab, agent-based models, and specific decision-making algorithms of blockchain technology adoptions in urban logistics and Smart City paradigms to fill the identified gaps in the literature. The main contribution of this study is to identify the research gaps in the analysis of the impact of blockchain on decarbonization in urban logistics.
  • Publication
    A Dataset for Explainable Sentiment Analysis in the German Automotive Industry
    ( 2023)
    Zielinski, Andrea
    ;
    Spolwind, Calvin
    ;
    Grimm, Anna orcid-logo
    ;
    Kroll, Henning orcid-logo
    While deep learning models have greatly improved the performance of many tasks related to sentiment analysis and classification, they are often criticized for being untrustworthy due to their black-box nature. As a result, numerous explainability techniques have been proposed to better understand the model predictions and to improve the deep learning models. In this work, we introduce InfoBarometer, the first benchmark for examining interpretable methods related to sentiment analysis in the German automotive sector based on online news. Each news article in our dataset is annotated with respect to the overall sentiment (i.e., positive, negative and neutral), the target of the sentiment (focusing on innovation-related topics such as e.g. electromobility) and the rationales, i.e., textual explanations for the sentiment label that can be leveraged during both training and evaluation. For this research, we compare different state-of-the-art approaches to perform sentiment analysis and observe that even models that perform very well in classification do not score high on explainability metrics like model plausibility and faithfulness. We calculated the polarity scores for the best method BERT and got a macro F1-score of 73.8. Moreover, we evaluated different interpretability algorithms (LIME, SHAP, Integrated Gradients, Saliency) based on explicitly marked rationales by human annotators quantitatively and qualitatively. Our experiments demonstrate that the textual explanations often do not agree with human interpretations, and rarely help to justify the models decision. However, global features provide useful insights to help uncover spurious features in the model and biases within the dataset. We intend to make our dataset public for other researchers.