Filters








124,406 Hits in 4.2 sec

SHAPE: Shifted Absolute Position Embedding for Transformers [article]

Shun Kiyono, Sosuke Kobayashi, Jun Suzuki, Kentaro Inui
2021 arXiv   pre-print
We investigate shifted absolute position embedding (SHAPE) to address both issues.  ...  Position representation is crucial for building position-aware representations in Transformers.  ...  Acknowledgements We thank the anonymous reviewers for their insightful comments. We thank Sho Takase for valuable discussions.  ... 
arXiv:2109.05644v1 fatcat:dkcbmo5jufgffktflcpbm6none

What I have learned from the research of "SHAPE: Shifted Absolute Position Embedding for Transformers"

Shun Kiyono
2022 Journal of Natural Language Processing  
該当箇所:https://github.com/butsugiri/shape/blob/main/onmt/modules/embeddings.py#L64-L75 3 当然これらはすべて Transformer の登場以後の研究である.  ...  4 同論文はその後 NeurIPS2021 に採択されたようである. 5 "position" で予稿集のタイトルを検索した. 6 5/17 締め切りのところ,5/5 の時点で投稿可能な状態の原稿が完成していたようである.残りの時間は全て原稿の クオリティ向上に注ぎ込まれた.  ... 
doi:10.5715/jnlp.29.248 fatcat:k2lqt2mfhzc6zl2ic2xihkriee

SHAPE: Shifted Absolute Position Embedding for Transformers

Shun Kiyono, Sosuke Kobayashi, Jun Suzuki, Kentaro Inui
2021 Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing   unpublished
We investigate shifted absolute position embedding (SHAPE) to address both issues.  ...  Position representation is crucial for building position-aware representations in Transformers.  ...  Acknowledgements We thank the anonymous reviewers for their insightful comments. We thank Sho Takase for valuable discussions.  ... 
doi:10.18653/v1/2021.emnlp-main.266 fatcat:fnwua4n6cjggjmtxjkptxgxpoq

Shape google

Alexander M. Bronstein, Michael M. Bronstein, Leonidas J. Guibas, Maks Ovsjanikov
2011 ACM Transactions on Graphics  
Our approach achieves state-of-the-art results on the SHREC 2010 large-scale shape retrieval benchmark.  ...  Finally, adopting metric learning approaches, we show that shapes can be efficiently represented as binary codes.  ...  Acknowledgment We are grateful to Zhouhui Lian and Umberto Castellani for providing the performance of their algorithms on the SHREC'10 benchmark, and to Giuseppe Patané for providing the FEM computation  ... 
doi:10.1145/1899404.1899405 fatcat:rwtead35svcmrplfxr3lfmtz6a

The shape of cubic fields [article]

Robert Hough
2019 arXiv   pre-print
We use the method of Shintani, as developed by Taniguchi and Thorne, to prove the quantitative equidistribution of the shape of cubic fields when the fields are ordered by discriminant.  ...  Let the real embeddings be σ 1 , ..., σ r 1 : K → R and the complex embeddings be σ r 1 +1 , ..., σ r 1 +r 2 : K → C.  ...  Note that, as σ(1) is present in the embedding for all K, σ(O) always has a short vector in this direction, relative to the volume.  ... 
arXiv:1705.06393v5 fatcat:v7k67tszsfdnxfntqlsjqguchu

Intuitive Shape Editing in Latent Space [article]

Tim Elsner, Moritz Ibing, Victor Czech, Julius Nehring-Wirxel, Leif Kobbelt
2022 arXiv   pre-print
The use of autoencoders for shape editing or generation through latent space manipulation suffers from unpredictable changes in the output shape.  ...  The control points on the surface that are part of the latent code of an object can then be freely moved, allowing for intuitive shape editing directly in latent space.  ...  For this, we apply a random shift of distance in [−0.03, 0.03] to each point, then de-and encode and measure the absolute mean difference to the input.  ... 
arXiv:2111.12488v3 fatcat:raq7j2ptmrgwhj3alh6sebitiu

End-to-end Lane Shape Prediction with Transformers [article]

Ruijin Liu, Zejian Yuan, Tie Liu, Zhiliang Xiong
2020 arXiv   pre-print
The lane shape model is formulated based on road structures and camera pose, providing physical interpretation for parameters of network output.  ...  To tackle these issues, we propose an end-to-end method that directly outputs parameters of a lane shape model, using a network built with a transformer to learn richer structures and context.  ...  The S and positional embedding E p are fed into the transformer encoder to output a representation sequence S e .  ... 
arXiv:2011.04233v2 fatcat:wryh4u5vtzaxdfbr7ivjaces3e

D-Former: A U-shaped Dilated Transformer for 3D Medical Image Segmentation [article]

Yixuan Wu, Kuanlun Liao, Jintai Chen, Jinhong Wang, Danny Z. Chen, Honghao Gao, Jian Wu
2022 arXiv   pre-print
Based on this design of Dilated Transformer, we construct a U-shaped encoder-decoder hierarchical architecture called D-Former for 3D medical image segmentation.  ...  Hence, recent work presented computer vision Transformer variants for medical image segmentation tasks and obtained promising performances.  ...  x i denotes the output feature maps embedded with position information.  ... 
arXiv:2201.00462v2 fatcat:qewt25275bbzzongwb3jztqtwu

Shape Dynamics. An Introduction [article]

Julian Barbour
2011 arXiv   pre-print
Shape dynamics is a completely background-independent universal framework of dynamical theories from which all absolute elements have been eliminated.  ...  For particles, only the variables that describe the shapes of the instantaneous particle configurations are dynamical.  ...  shape s 2 into the unique position in its orbit at which (8) is minimized (for given position of s 1 ).  ... 
arXiv:1105.0183v1 fatcat:ss73vxld4bdt5lcpgkkchicrhe

Investigating transformers in the decomposition of polygonal shapes as point collections [article]

Andrea Alfieri, Yancong Lin, Jan C. van Gemert
2021 arXiv   pre-print
We provide evidence on the importance of natural orders for Transformers, and show the benefit of decomposing complex polygons into collections of points in an auto-regressive manner.  ...  mostly explored upon this difference on sequential tasks in NLP, we study the difference between auto-regressive and parallel prediction on visual set prediction tasks, and in particular on polygonal shapes  ...  The feature map is supplemented with fixed positional embeddings before the Transformer encoder.  ... 
arXiv:2108.07533v1 fatcat:f75zpuposncmnplitsqpmud7za

Design and experimental testing of an adaptive shape-morphing tensegrity structure, with frequency self-tuning capabilities, using shape-memory alloys

Filipe Amarante dos Santos, André Rodrigues, Andrea Micheletti
2015 Smart materials and structures (Print)  
The implementation of a short-time-Fourier-transform control algorithm allows for the correct identification of the dominant input frequency, associated with the dynamic excitation.  ...  This actuation principle also takes advantage of the variation of the elastic modulus of shape-memory alloys associated with the martensitic transformation.  ...  We also wish to thank an anonymous reviewer for several corrections and useful suggestions.  ... 
doi:10.1088/0964-1726/24/10/105008 fatcat:bityowf4gfbohkbrpip25th76q

Transformation of Embedded Shape Memory Alloy Ribbons

K. D. Jonnalagadda, N. R. Sottos, M. A. Qidwai, D. C. Lagoudas
1998 Journal of Intelligent Materials Systems and Structures  
Shape memory alloy (SMA) wires can be embedded in a host material to alter the stiffness or modal response and provide vibration control.  ...  Although there has been a significant amount of research dedicated to characterizing and modeling the response of SMA alone, little research has focused on the transformation behavior of embedded SMAs.  ...  This shows the shifting of stress concentration from the debonded area towards the free edge, which clearly points out the reason for less transformation in that region than the debonded case for the same  ... 
doi:10.1177/1045389x9800900506 fatcat:gsb7otakpbewbh4p42kxrbrviy

Transformation of embedded shape memory alloy ribbons

Krishna D. Jonnalagadda, Nancy R. Sottos, Muhammad A. Qidwai, Dimitris C. Lagoudas, Vasundara V. Varadan, Jagdish Chandra
1997 Smart Structures and Materials 1997: Mathematics and Control in Smart Structures  
Shape memory alloy (SMA) wires can be embedded in a host material to alter the stiffness or modal response and provide vibration control.  ...  Although there has been a significant amount of research dedicated to characterizing and modeling the response of SMA alone, little research has focused on the transformation behavior of embedded SMAs.  ...  This shows the shifting of stress concentration from the debonded area towards the free edge, which clearly points out the reason for less transformation in that region than the debonded case for the same  ... 
doi:10.1117/12.276543 fatcat:7yie6db53vbxrf245mipdmdqtm

SPH hydrocodes can be stabilized with shape-shifting

D.L. Hicks, L.M. Liebrock
1999 Computers and Mathematics with Applications  
SPH can also be stabilized by shifting the shape of W to change the sign of W'.  ...  Shape-shifting can be used to reduce the amount of smoothing needed. Section 8, The Shape-Shifter Stability Theorem, discusses shape-shifting to remove the tension instability in SPH.  ...  shape-shifter for tension.Note the negative and positive regions for W":Wt'(v,h,~) < O, for ~L < [Vl < ~m,(45)W"(v, h, ~) > 0,for ~m < Ivl < ~, ) ff there is a YM E [0, 1] such that G'(yM) = O, then  ... 
doi:10.1016/s0898-1221(99)00210-2 fatcat:rkeg7sbyjjhhpcatcybuxp55ci

Shape Palindromes: Analysis of Intrinsic Symmetries in 2D Articulated Shapes [chapter]

Amit Hooda, Michael M. Bronstein, Alexander M. Bronstein, Radu P. Horaud
2012 Lecture Notes in Computer Science  
For that purpose, we show two efficient computational methods: one based on Fourier analysis, and another on dynamic programming.  ...  Analysis of intrinsic symmetries of non-rigid and articulated shapes is an important problem in pattern recognition with numerous applications ranging from medicine to computational aesthetics.  ...  To account for the noise, we only look at a neighborhood of the A k with maximum the absolute value. Reflection Symmetry.  ... 
doi:10.1007/978-3-642-24785-9_56 fatcat:jqyr2tsbojgx3idshmkx5j7hhm
« Previous Showing results 1 — 15 out of 124,406 results