4 Hits in 2.8 sec

Fixes That Fail: Self-Defeating Improvements in Machine-Learning Systems [article]

Ruihan Wu, Chuan Guo, Awni Hannun, Laurens van der Maaten
2021 arXiv   pre-print
Models in these systems are often developed and trained independently, which raises an obvious concern: Can improving a machine-learning model make the overall system worse?  ...  We also present the results of experiments which show that self-defeating improvements emerge in a realistic stereo-based detection system for cars and pedestrians.  ...  Acknowledgements The authors thank Yan Wang, Yurong You, and Ari Morcos for helpful discussions and advice.  ... 
arXiv:2103.11766v2 fatcat:3srqii5mnzb7fh4h2oyrgj47y4

One Model to Synthesize Them All: Multi-contrast Multi-scale Transformer for Missing Data Imputation [article]

Jiang Liu, Srivathsa Pasumarthi, Ben Duffy, Enhao Gong, Greg Zaharchuk, Keshav Datta
2022 arXiv   pre-print
Thanks to the proposed multi-contrast Swin Transformer blocks, it can efficiently capture intra- and inter-contrast dependencies for accurate image synthesis.  ...  In this paper, we formulate missing data imputation as a sequence-to-sequence learning problem and propose a multi-contrast multi-scale Transformer (MMT), which can take any subset of input contrasts and  ...  The green overlay is the Tumor Core mask segmented on the respective images using an automatic tumor segmentation model [51] .  ... 
arXiv:2204.13738v1 fatcat:kv5bmx5qbneexpfjknmw3oc75i

Pre-Proceedings of the International Colloquium on Graph and Model Transformation – On the occasion of the 65th birthday of Hartmut Ehrig (GraMoT 2010) [article]

(:Unkn) Unknown, Technische Universität Berlin, Technische Universität Berlin, Claudia Ermel, Hartmut Ehrig, Fernando Orejas, Gabriele Taentzer
Padberg et al. for formal modeling and analysis of flexible processes in communication based systems, like mobile ad-hoc networks and Skype. iv Pre-Proceedings GraMoT 2010 and visual techniques and environments  ...  Finally, an integration of Petri nets and high-level replacement systems, known as reconfigurable Petri nets, is used in the third paper by K. Hoffmann, J.  ...  Acknowledgements: We want to thank Fabio Gadducci, Alberto Lluch Lafuente, Daniele Terreni and Liang Zhao for many interesting discussions and exchange of ideas regarding the graphical encoding of CaSPiS  ... 
doi:10.14279/depositonce-10237 fatcat:44oa7p3275fhpdl4k5kdjr7euy

Simulation and Animation of Visual Languages based on Typed Algebraic Graph Transformation [article]

Claudia Ermel, Technische Universität Berlin, Technische Universität Berlin, Hartmut Ehrig
Thus, they provide better insights of model behaviour and lead to an earlier detection of inconsistencies and possible missing requirements in the model.  ...  A simulation specification is mapped to an animation view by so-called simulation-to-animation model-and-rule transformation (S2A transformation).  ...  are generated by the rule are highlighted by green color.  ... 
doi:10.14279/depositonce-1417 fatcat:7bexnsuegvbohifihll6fzlox4