12 research outputs found
TRANSFORMER TECHNIQUES FOR HUMAN ACTION RECOGNITION AND LOCALIZATION
No full textPh.DDOCTOR OF PHILOSOPHY (CDE-ENG
Melanie Kühn, Die Bekämpfung der Steuerumgehung durch allgemeine Antimissbrauchsvorschriften in Hongkong, Singapur und der Volksrepublik China – Auf dem Weg zur Konvergenz auf internationaler Ebene?, in: Feld/Reimer/Waldhoff (eds.), Steuerwissenschaftliche Schriften 86, 1st Edition, Baden-Baden 2024
No full textDie Dissertation von Melanie Kühn, die im Sommersemester 2023 von der Juristischen Fakultät der Ludwig-Maximilians-Universität München angenommen wurde, befasst sich mit dem hochaktuellen und komplexen Thema der Bekämpfung der Steuerumgehung durch allgemeine Antimissbrauchsvorschriften (general anti-avoidance rules, „GAAR“) in den ostasiatischen Jurisdiktionen Hongkong, Singapur und der VR China.
Die Arbeit, die Kühn während ihrer Zeit als wissenschaftliche Mitarbeiterin am Max-Planck-Institut für Steuerrecht und Öffentliche Finanzen verfasste, bietet nicht nur eine tiefgehende Analyse der rechtlichen Rahmenbedingungen und der praktischen Anwendung von GAAR in diesen drei unterschiedlichen Rechtssystemen. Sie legt zudem den Fokus auf die Untersuchung, ob diese ostasiatischen Jurisdiktionen ähnliche Divergenzprobleme im Steuerrecht aufweisen wie andere Länder, und darauf, ob das von Christine Osterloh-Konrad entwickelte Zwei-Stufen-Modell zur Identifizierung und Kategorisierung von Gestaltungsmissbrauch universell anwendbar ist.The dissertation of Melanie Kühn, accepted by the Law Faculty of the Ludwig Maximilian University of Munich in the summer of 2023, deals with the highly topical and complex issue of combating tax avoidance through general anti-avoidance rules ("GAAR"), as specific to the East Asian jurisdictions of Hong Kong, Singapore, and the People\u27s Republic of China.
The work, written by Kühn during her time as a research assistant at the Max Planck Institute for Tax Law and Public Finance, offers an in-depth analysis of the legal framework and practical application of GAAR in the three legal systems. Moreover, Kühn examines whether these East Asian jurisdictions are experiencing tax law divergence problems similar to those in other countries; she also considers the universal applicability of the two-stage model developed by Christine Osterloh-Konrad for identifying and categorizing abusive arrangements
Remediation of Copper Contaminated Kaolin by Electrokinetics Coupled with Permeable Reactive Barrier
No full textAbstractElectrokinetics is an in situ soil remediation technique by which the flow direction of the pollutants can be controlled and the soil with low permeability can be treated. In this study, the remediation of copper contaminated kaolin by electrokinetic process coupled with activated carbon permeable reactive barrier (PRB) was investigated. The experimental results showed that the integration of PRB with electrokinetics successfully removed copper from kaolin with pH control of the catholyte. The average removal rate reached the highest of 96.60% when the initial Cu2+ concentration was 2000mg/kg. Compared to the electrokinetic process without PRB, the application of the coupled system could reduce the pollution of the electrolyte
SALIENCY DETECTION VIA GLOBAL-OBJECT-SEED-GUIDED CELLULAR AUTOMATA
No full textImage saliency detection has attracted much attention in recent years, while several challenging problems are still unsolved, such as inaccurate saliency detection in complex scenes and suppressing salient objects near image borders. In this paper, a novel algorithm is proposed to solve these problems. Firstly, we collect background seeds from image borders by boundary information and construct a background based saliency map via low level features. Then, a novel propagation mechanism named global-object-seed-guided Cellular Automata model is builded. Cellular Automata exploits the intrinsic relevance of similar regions through interactions with neighbors, and global object seeds reduce the difference between dissimilar adjacent regions in the whole salient object. Experimental results on public benchmark datasets demonstrate the superiority of the proposed algorithm over ten state-of-the-art saliency models.CPCI-S(ISTP)[email protected]; [email protected]; [email protected]
Datasets for replicating the paper "Raman Spectrum Matching with Contrastive Representation Learning"
No full textDatasets Mineral and Organic for replicating the paper "Raman spectrum matching with contrastive learning".
The detailed instruction about how to use the dataset, please visit our Github Repository.
To use these two datasets, please cite:
@inproceedings{Lafuente2016ThePO,
title={The power of databases: The RRUFF project},
author={B. Lafuente and R. Downs and Hexiong Yang and N. Stone},
booktitle = {Highlights in Mineralogical Crystallography},
year={2016}
}
@article{organic_dataset,
author = {Zhang, Rui and Xie, Huimin and Cai, Shuning and Hu, Yong and Liu, Guo-kun and Hong, Wenjing and Tian, Zhong-qun},
title = {Transfer-learning-based Raman spectra identification},
journal = {Journal of Raman Spectroscopy},
volume = {51},
number = {1},
pages = {176-186},
keywords = {deep learning, Raman spectroscopy, transfer learning},
year = {2020}
}
@Article{D2AN00403H,
author ="Li, Bo and Schmidt, Mikkel N. and Alstrøm, Tommy S.",
title ="Raman spectrum matching with contrastive representation learning",
journal ="Analyst",
year ="2022",
volume ="147",
issue ="10",
pages ="2238-2246",
publisher ="The Royal Society of Chemistry",
doi ="https://doi.org/10.1039/d2an00403h",
url ="http://dx.doi.org/10.1039/D2AN00403H",
}</p
Revisiting Vision Transformer from the View of Path Ensemble
No full textVision Transformers (ViTs) are normally regarded as a stack of transformer
layers. In this work, we propose a novel view of ViTs showing that they can be
seen as ensemble networks containing multiple parallel paths with different
lengths. Specifically, we equivalently transform the traditional cascade of
multi-head self-attention (MSA) and feed-forward network (FFN) into three
parallel paths in each transformer layer. Then, we utilize the identity
connection in our new transformer form and further transform the ViT into an
explicit multi-path ensemble network. From the new perspective, these paths
perform two functions: the first is to provide the feature for the classifier
directly, and the second is to provide the lower-level feature representation
for subsequent longer paths. We investigate the influence of each path for the
final prediction and discover that some paths even pull down the performance.
Therefore, we propose the path pruning and EnsembleScale skills for
improvement, which cut out the underperforming paths and re-weight the ensemble
components, respectively, to optimize the path combination and make the short
paths focus on providing high-quality representation for subsequent paths. We
also demonstrate that our path combination strategies can help ViTs go deeper
and act as high-pass filters to filter out partial low-frequency signals. To
further enhance the representation of paths served for subsequent paths,
self-distillation is applied to transfer knowledge from the long paths to the
short paths. This work calls for more future research to explain and design
ViTs from new perspectives.Comment: Accepted by ICCV 2023, oral presentatio
Surfactant-free synthesis of hierarchical niobic acid microflowers assembled from ultrathin nanosheets with efficient photoactivities
No full textHierarchical niobic acid (Nb2O5.nH(2)O) microflowers are synthesized by a surfactant-free hydrothermal approach. The three-dimensional microflowers are assembled from two-dimensional ultrathin nanosheets with thickness of similar to 5 nm. Using rhodamine B as a probe, the Nb2O5.nH(2)O microflowers exhibit high photocatalytic activity under UV light irradiation. Furthermore, the Nb2O5.nH(2)O microflowers are easily converted to niobium pentoxide without significant structural alteration. (C) 2016 Elsevier B.V. All rights reserved.National Research Foundation (NRF), Prime Minister's Office, Singapore under its Campus for Research Excellence and Technological Enterprise (CREATE) programme; Singapore-Peking University Research Centre for a Sustainable Low-Carbon Future [R143-001-205-592]; MOE [R143-000-542-112]SCI(E)ARTICLE514-52239
Mix-of-Show: Decentralized Low-Rank Adaptation for Multi-Concept Customization of Diffusion Models
No full textPublic large-scale text-to-image diffusion models, such as Stable Diffusion,
have gained significant attention from the community. These models can be
easily customized for new concepts using low-rank adaptations (LoRAs). However,
the utilization of multiple concept LoRAs to jointly support multiple
customized concepts presents a challenge. We refer to this scenario as
decentralized multi-concept customization, which involves single-client concept
tuning and center-node concept fusion. In this paper, we propose a new
framework called Mix-of-Show that addresses the challenges of decentralized
multi-concept customization, including concept conflicts resulting from
existing single-client LoRA tuning and identity loss during model fusion.
Mix-of-Show adopts an embedding-decomposed LoRA (ED-LoRA) for single-client
tuning and gradient fusion for the center node to preserve the in-domain
essence of single concepts and support theoretically limitless concept fusion.
Additionally, we introduce regionally controllable sampling, which extends
spatially controllable sampling (e.g., ControlNet and T2I-Adaptor) to address
attribute binding and missing object problems in multi-concept sampling.
Extensive experiments demonstrate that Mix-of-Show is capable of composing
multiple customized concepts with high fidelity, including characters, objects,
and scenes
