Fakultät Informatik und Mathematik
Refine
Year of publication
- 2019 (5) (remove)
Document Type
Has Fulltext
- yes (5) (remove)
Is part of the Bibliography
- no (5)
Keywords
- Maschinelles Lernen (3)
- Adam optimizer (2)
- Convergence (2)
- Konvergenz 〈Informationstechnik〉 (2)
- Neuronales Netz (2)
- Optimierungsalgorithmus (2)
- Backsourcing (1)
- Computer algebra (1)
- Computeralgebra (1)
- Computerunterstütztes Verfahren (1)
Institute
Begutachtungsstatus
- peer-reviewed (5) (remove)
Adaptive Moment Estimation (Adam) is a very popular training algorithm for deep neural networks, implemented in many machine learning frameworks. To the best of the authors knowledge no complete convergence analysis exists for Adam. The contribution of this paper is a method for the local convergence analysis in batch mode for a deterministic fixed training set, which gives necessary conditions for the hyperparameters of the Adam algorithm. Due to the local nature of the arguments the objective function can be non-convex but must be at least twice continuously differentiable.
Computer-aided diagnosis using deep learning in the evaluation of early oesophageal adenocarcinoma
(2019)
Computer-aided diagnosis using deep learning (CAD-DL) may be an instrument to improve endoscopic assessment of Barrett’s oesophagus
(BE) and early oesophageal adenocarcinoma (EAC). Based on still images from two databases, the diagnosis of EAC by CAD-DL reached sensitivities/specificities of 97%/88% (Augsburg data) and 92%/100% (Medical Image Computing and Computer-Assisted Intervention [MICCAI]
data) for white light (WL) images and 94%/80% for narrow band images (NBI) (Augsburg data), respectively. Tumour margins delineated by
experts into images were detected satisfactorily with a Dice coefficient (D) of 0.72. This could be a first step towards CAD-DL for BE assessment. If developed further, it could become a useful
adjunctive tool for patient management.
Beim Edelmetallcontrolling von Infineon sind Fehler nur mit hohem Aufwand analysierbar und es bewirkt widersprechende Berechnungshinweise. Alle vier Perspektiven des Edelmetallcontrollings werden verbessert und als einzige Datenquelle wird die von SAP festgesetzt.
Weitere Verbesserungen werden durch die Umwandlung einer monatlichen
Kursänderungserhebung in eine jährliche, durch die Verwendung des Marktwerts für den Bestandswert sowie durch Ermittlung eines
Normwerts für eine jährliche Recyclingquote erzielt.
To prepare their IT landscape for future business challenges, companies are changing their IT sourcing arrangements by using selective sourcing approaches as well as multi-sourcing with more but smaller sourcing contracts. Companies therefore have to reconsider and re-evaluate their IT sourcing setup more frequently. Collecting data from 251 global experts, we empirically tested the effect of service quality, relationship quality, and switching costs on IT sourcing decisions using partial least squares (PLS) analysis. Drawing on previously conducted expert interviews, our model extends previous studies and introduces a decision maker’s sourcing preferences as a not yet examined moderator on IT sourcing decisions. This allows us to investigate the influence of the decision maker’s beliefs on the decision process. We were able to confirm the negative effect of switching costs on a decision in favor of backsourcing, however we could not find significant support for the remaining hypotheses. We further discuss potential reasons for our findings and suggest future research opportunities based on our contribution.
One of the most popular training algorithms for deep neural networks is the Adaptive Moment Estimation (Adam) introduced by Kingma and Ba. Despite its success in many applications there is no satisfactory convergence analysis: only local convergence can be shown for batch mode under some restrictions on the hyperparameters, counterexamples exist for incremental mode. Recent results show that for simple quadratic objective functions limit cycles of period 2 exist in batch mode, but only for atypical hyperparameters, and only for the algorithm without bias correction. We extend the convergence analysis to all choices of the hyperparameters for quadratic functions. This finally answers the question of convergence for Adam in batch mode to the negative. We analyze the stability of these limit cycles and relate our analysis to other results where approximate convergence was shown, but under the additional assumption of bounded gradients which does not apply to quadratic functions. The investigation heavily relies on the use of computer algebra due to the complexity of the equations.