Logo des Repositoriums

Auflistung nach Schlagwort "Deep learning"

1 - 10 von 11
  • Zeitschriftenartikel
    A Novel Business Process Prediction Model Using a Deep Learning Method
    (Business & Information Systems Engineering: Vol. 62, No. 2, 2020) Mehdiyev, Nijat; Evermann, Joerg; Fettke, Peter
    The ability to proactively monitor business processes is a main competitive differentiator for firms. Process execution logs generated by process aware information systems help to make process specific predictions for enabling a proactive situational awareness. The goal of the proposed approach is to predict the next process event from the completed activities of the running process instance, based on the execution log data from previously completed process instances. By predicting process events, companies can initiate timely interventions to address undesired deviations from the desired workflow. The paper proposes a multi-stage deep learning approach that formulates the next event prediction problem as a classification problem. Following a feature pre-processing stage with n-grams and feature hashing, a deep learning model consisting of an unsupervised pre-training component with stacked autoencoders and a supervised fine-tuning component is applied. Experiments on a variety of business process log datasets show that the multi-stage deep learning approach provides promising results. The study also compared the results to existing deep recurrent neural networks and conventional classification approaches. Furthermore, the paper addresses the identification of suitable hyperparameters for the proposed approach, and the handling of the imbalanced nature of business process event datasets.
  • Zeitschriftenartikel
    Automatic Classification of Bloodstains with Deep Learning Methods
    (KI - Künstliche Intelligenz: Vol. 36, No. 2, 2022) Bergman, Tommy; Klöden, Martin; Dreßler, Jan; Labudde, Dirk
    The classification of detected bloodstains into predetermined categories is a crucial component of the so-called bloodstain pattern analysis. As in other forensic disciplines, deep learning methods may help to reduce human subjectivity within this process, may increase the classification accuracy, shorten the calculation time and thus, enable high-throughput analysis. In this work, an approach is presented in which a convolutional neural network (Inception v3) was trained from 965 drip stains (passive origin) and 1595 blood spatters (active origin). The trained CNN was evaluated with a test data set consisting of 366 images of drip stains and blood spatters. The success rate was 99.73% which suggests that neural networks could also be used to automatically classify other classes of bloodstain patterns to speed up the investigation process in the future.
  • Zeitschriftenartikel
    Beyond Manual Tuning of Hyperparameters
    (KI - Künstliche Intelligenz: Vol. 29, No. 4, 2015) Hutter, Frank; Lücke, Jörg; Schmidt-Thieme, Lars
    The success of hand-crafted machine learning systems in many applications raises the question of making machine learning algorithms more autonomous, i.e., to reduce the requirement of expert input to a minimum. We discuss two strategies towards this goal: (1) automated optimization of hyperparameters (including mechanisms for feature selection, preprocessing, model selection, etc) and (2) the development of algorithms with reduced sets of hyperparameters. Since many research directions (e.g., deep learning), show a tendency towards increasingly complex algorithms with more and more hyperparamters, the demand for both of these strategies continuously increases. We review recent hyperparameter optimization methods and discuss data-driven approaches to avoid the introduction of hyperparameters using unsupervised learning. We end in discussing how these complementary strategies can work hand-in-hand, representing a very promising approach towards autonomous machine learning.
  • Zeitschriftenartikel
    Continuous Training and Deployment of Deep Learning Models
    (Datenbank-Spektrum: Vol. 21, No. 3, 2021) Prapas, Ioannis; Derakhshan, Behrouz; Mahdiraji, Alireza Rezaei; Markl, Volker
    Deep Learning (DL) has consistently surpassed other Machine Learning methods and achieved state-of-the-art performance in multiple cases. Several modern applications like financial and recommender systems require models that are constantly updated with fresh data. The prominent approach for keeping a DL model fresh is to trigger full retraining from scratch when enough new data are available. However, retraining large and complex DL models is time-consuming and compute-intensive. This makes full retraining costly, wasteful, and slow. In this paper, we present an approach to continuously train and deploy DL models. First, we enable continuous training through proactive training that combines samples of historical data with new streaming data. Second, we enable continuous deployment through gradient sparsification that allows us to send a small percentage of the model updates per training iteration. Our experimental results with LeNet5 on MNIST and modern DL models on CIFAR-10 show that proactive training keeps models fresh with comparable—if not superior—performance to full retraining at a fraction of the time. Combined with gradient sparsification, sparse proactive training enables very fast updates of a deployed model with arbitrarily large sparsity, reducing communication per iteration up to four orders of magnitude, with minimal—if any—losses in model quality. Sparse training, however, comes at a price; it incurs overhead on the training that depends on the size of the model and increases the training time by factors ranging from 1.25 to 3 in our experiments. Arguably, a small price to pay for successfully enabling the continuous training and deployment of large DL models.
  • Konferenzbeitrag
    A Deep Learning-based Approach for Banana Leaf Diseases Classification
    (Datenbanksysteme für Business, Technologie und Web (BTW 2017) - Workshopband, 2017) Amara, Jihen; Bouaziz, Bassem; Algergawy, Alsayed
    Plant diseases are important factors as they result in serious reduction in quality and quantity of agriculture products. Therefore, early detection and diagnosis of these diseases are important. To this end, we propose a deep learning-based approach that automates the process of classifying ba- nana leaves diseases. In particular, we make use of the LeNet architecture as a convolutional neural network to classify image data sets. The preliminary results demonstrate the effectiveness of the proposed approach even under challenging conditions such as illumination, complex background, different resolution, size, pose, and orientation of real scene images.
  • Zeitschriftenartikel
    Gone in 30 days! Predictions for car import planning
    (it - Information Technology: Vol. 60, No. 4, 2018) Lacic, Emanuel; Traub, Matthias; Duricic, Tomislav; Haslauer, Eva; Lex, Elisabeth
    A challenge for importers in the automobile industry is adjusting to rapidly changing market demands. In this work, we describe a practical study of car import planning based on the monthly car registrations in Austria. We model the task as a data driven forecasting problem and we implement four different prediction approaches. One utilizes a seasonal ARIMA model, while the other is based on LSTM-RNN and both compared to a linear and seasonal baselines. In our experiments, we evaluate the 33 different brands by predicting the number of registrations for the next month and for the year to come.
  • Zeitschriftenartikel
    Machine Learning in Business Process Monitoring: A Comparison of Deep Learning and Classical Approaches Used for Outcome Prediction
    (Business & Information Systems Engineering: Vol. 63, No. 3, 2021) Kratsch, Wolfgang; Manderscheid, Jonas; Röglinger, Maximilian; Seyfried, Johannes
    Predictive process monitoring aims at forecasting the behavior, performance, and outcomes of business processes at runtime. It helps identify problems before they occur and re-allocate resources before they are wasted. Although deep learning (DL) has yielded breakthroughs, most existing approaches build on classical machine learning (ML) techniques, particularly when it comes to outcome-oriented predictive process monitoring. This circumstance reflects a lack of understanding about which event log properties facilitate the use of DL techniques. To address this gap, the authors compared the performance of DL (i.e., simple feedforward deep neural networks and long short term memory networks) and ML techniques (i.e., random forests and support vector machines) based on five publicly available event logs. It could be observed that DL generally outperforms classical ML techniques. Moreover, three specific propositions could be inferred from further observations: First, the outperformance of DL techniques is particularly strong for logs with a high variant-to-instance ratio (i.e., many non-standard cases). Second, DL techniques perform more stably in case of imbalanced target variables, especially for logs with a high event-to-activity ratio (i.e., many loops in the control flow). Third, logs with a high activity-to-instance payload ratio (i.e., input data is predominantly generated at runtime) call for the application of long short term memory networks. Due to the purposive sampling of event logs and techniques, these findings also hold for logs outside this study.
  • Zeitschriftenartikel
    Mit Computer Vision zur automatisierten Qualitätssicherung in der industriellen Fertigung: Eine Fallstudie zur Klassifizierung von Fehlern in Solarzellen mittels Elektrolumineszenz-Bildern
    (HMD Praxis der Wirtschaftsinformatik: Vol. 58, No. 2, 2021) Zschech, Patrick; Sager, Christoph; Siebers, Philipp; Pertermann, Maik
    Die Qualitätssicherung bei der Produktion von Solarzellen ist ein entscheidender Faktor, um langfristige Leistungsgarantien auf Solarpanels gewähren zu können. Die vorliegende Arbeit leistet hierzu einen Beitrag zur automatisierten Fehlererkennung auf Wafern, indem Elektrolumineszenz-Bilder eines realen Herstellungsszenarios mithilfe von verschiedenen Computer-Vision-Modellen klassifiziert werden. Die Herausforderung besteht hierbei nicht nur darin, defekte Wafer von funktionsfähigen zu separieren, sondern gleichzeitig auch zwischen spezifischen Fehlerarten zu unterscheiden, während geringe Inferenzzeiten sicherzustellen sind. Zu diesem Zweck werden neben einfachen statistischen Modellen verschiedene Deep-Learning-Architekturen auf Basis von Convolutional Neural Networks (CNNs) verprobt und miteinander vergleichen. Ziel der Arbeit ist es, verschiedene Klassifizierungsansätze unterschiedlicher Komplexität zu testen und auf ihre praktische Einsatzfähigkeit unter realen Bedingungen zu untersuchen. Die Fallstudie zeigt, dass je nach Situation unterschiedliche Modelle ihre Existenzberechtigung haben und in Kombination sehr gute Ergebnisse erzielen. So lassen sich bereits mit statistischen Modellen und einfachen CNN-Varianten zuverlässige Aussagen mit Genauigkeiten von über 99 % bei Fehlertypen einfacher bis mittlerer Erkennbarkeit realisieren. Werden die Fehlerbilder demgegenüber diffuser und soll die Nachvollziehbarkeit der Ergebnisse durch positionsgenaue Lokalisierung von Fehlerobjekten gewährleistet werden, sind fortgeschrittenere Ansätze auf Basis sogenannter Region-Proposal-Netzwerke erforderlich, die allerdings auch mit einem erhöhten Labeling-Aufwand beim Annotieren der Fehlerobjekte einhergehen. Da die Umsetzung sämtlicher Modelle ausschließlich auf Open Source Tools wie zum Beispiel TensorFlow, Keras und OpenCV basiert, demonstriert die Fallstudie zudem, welche Möglichkeiten durch frei verfügbare Lösungen im Bereich von Computer Vision geboten werden. Quality assurance in the production of solar cells is a decisive factor for long-term performance guarantees on solar panels. This work contributes to this area in developing computer vision models to automatically detect defects on wafers by classifying electroluminescence images from a real manufacturing scenario. The challenge is not only to separate defective wafers from flawless ones but also to distinguish between specific types of defects while ensuring low inference times. For this purpose, simple statistical models, as well as different kinds of deep learning architectures based on convolutional neural networks (CNNs), are tested and compared with each other. Therefore, this work aims to evaluate multiple classification approaches of varying complexity levels while examining their practical applicability under real industrial conditions. The case study shows that all models have their right to exist and achieve excellent results in combination. While statistical models and simple CNNs provide reliable statements with accuracies up to 99% for defect types of simple to medium detectability, more advanced approaches based on region proposal networks are required once the defect images become more diffuse. The more advanced approaches allow a precise object localization of defects; however, they are also associated with increased labeling effort when annotating wafer images. Since the implementation of all models is based exclusively on open source tools such as TensorFlow, Keras, and OpenCV, the case study also demonstrates the possibilities offered by freely accessible solutions in the field of computer vision.
  • Zeitschriftenartikel
    Towards Explanatory Interactive Image Captioning Using Top-Down and Bottom-Up Features, Beam Search and Re-ranking
    (KI - Künstliche Intelligenz: Vol. 34, No. 4, 2020) Biswas, Rajarshi; Barz, Michael; Sonntag, Daniel
    Image captioning is a challenging multimodal task. Significant improvements could be obtained by deep learning. Yet, captions generated by humans are still considered better, which makes it an interesting application for interactive machine learning and explainable artificial intelligence methods. In this work, we aim at improving the performance and explainability of the state-of-the-art method Show, Attend and Tell by augmenting their attention mechanism using additional bottom-up features. We compute visual attention on the joint embedding space formed by the union of high-level features and the low-level features obtained from the object specific salient regions of the input image. We embed the content of bounding boxes from a pre-trained Mask R-CNN model. This delivers state-of-the-art performance, while it provides explanatory features. Further, we discuss how interactive model improvement can be realized through re-ranking caption candidates using beam search decoders and explanatory features. We show that interactive re-ranking of beam search candidates has the potential to outperform the state-of-the-art in image captioning.
  • Konferenzbeitrag
    Vudenc: Vulnerability Detection with Deep Learning on a Natural Codebase for Python - Summary
    (Software Engineering 2023, 2023) Wartschinski, Laura; Noller, Yannic; Vogel, Thomas; Kehrer, Timo; Grunske, Lars
    In this extended abstract, we summarize our work on Vudenc published in the journal Information and Software Technology (IST) in 2022 [Wa22]. Vudenc uses deep learning to learn features of vulnerable code from a real-world Python codebase and a network of long-short-term memory cells (LSTM) is then used to detect vulnerabilities in code at a fine-grained level.