A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
Filters
Vision Transformer Slimming: Multi-Dimension Searching in Continuous Optimization Space
[article]
2022
arXiv
pre-print
Preserved Fulltext
Other Versions
This paper explores the feasibility of finding an optimal sub-model from a vision transformer and introduces a pure vision transformer slimming (ViT-Slim) framework. ...
Our method is based on a learnable and unified ℓ_1 sparsity constraint with pre-defined factors to reflect the global importance in the continuous searching space of different dimensions. ...
The last advantage is that we can search for a finergrained architecture such as the different dimensionalities in different self-attention heads, as our search space is continuous in them. ...
arXiv:2201.00814v2
fatcat:2ia77hhmfndb7blaqry4sdjejm
Super Vision Transformer
[article]
2022
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
Other Versions
We attempt to reduce the computational costs in vision transformers (ViTs), which increase quadratically in the token number. ...
Also, our SuperViT significantly outperforms existing studies on efficient vision transformers. ...
ViT-Slim [5] searches for a sub-transformer network across three dimensions of input tokens, MHSA and MLP modules with a 1 -regularized soft mask to indicate the global importance of dimensions, just ...
arXiv:2205.11397v2
fatcat:icjeohab7ncijahldaw6vvpsym
A Survey on Vision Transformer
[article]
2021
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2021; you can also visit the original URL.
The file type is application/pdf.
Other Versions
Furthermore, we also take a brief look at the self-attention mechanism in computer vision, as it is the base component in transformer. ...
In this paper, we review these vision transformer models by categorizing them in different tasks and analyzing their advantages and disadvantages. ...
[290] extended the network slimming approach [147] to vision transformers for reducing the dimensions of linear projections in both FFN and attention modules. ...
arXiv:2012.12556v4
fatcat:ldtbdgy6tbdttfqzhzml7n577m
TinyViT: Fast Pretraining Distillation for Small Vision Transformers
[article]
2022
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
Vision transformer (ViT) recently has drawn great attention in computer vision due to its remarkable model capability. ...
To alleviate this issue, we propose TinyViT, a new family of tiny and efficient small vision transformers pretrained on large-scale datasets with our proposed fast distillation framework. ...
It may help both the manual design and the search space design for efficient small vision transformers. 1) For small vision transformers, it improves the accuracy when replacing the transformer block in ...
arXiv:2207.10666v1
fatcat:3xkyjmockvhmbltkj3jhsur2sa
Chasing Sparsity in Vision Transformers: An End-to-End Exploration
[article]
2021
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2021; you can also visit the original URL.
The file type is application/pdf.
Other Versions
Vision transformers (ViTs) have recently received explosive popularity, but their enormous model sizes and training costs remain daunting. ...
Our approach jointly optimizes model parameters and explores connectivity throughout training, ending up with one sparse network as the final output. ...
Acknowledgment Z.W. is in part supported by an NSF RTML project (#2053279). ...
arXiv:2106.04533v3
fatcat:5fua73defbfgfeln7ifd6g56ce
EdgeViTs: Competing Light-weight CNNs on Mobile Devices with Vision Transformers
[article]
2022
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
Other Versions
Self-attention based models such as vision transformers (ViTs) have emerged as a very competitive architecture alternative to convolutional neural networks (CNNs) in computer vision. ...
Specifically, we show that our models are Pareto-optimal when both accuracy-latency and accuracy-energy trade-offs are considered, achieving strict dominance over other ViTs in almost all cases and competing ...
Vision transformers. ViTs [10] quickly popularize transformer-based architectures for computer vision. ...
arXiv:2205.03436v2
fatcat:dybmfnlw45hepb5hv2ilxeiuam
Non-terrestrial Communications Assisted by Reconfigurable Intelligent Surfaces
[article]
2021
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2021; you can also visit the original URL.
The file type is application/pdf.
Then we overview the literature related to RANTNs from the perspectives of performance analysis and optimization, followed by the widely used methodologies. ...
In combination with next-generation communication technologies, the advanced technologies in RANTNs are discussed. ...
To handle the case of continuous state and action spaces, policy gradient methods have been proposed. ...
arXiv:2109.00876v1
fatcat:gukzusg4rzeevcfzdl2ff7yxdm
Artificial Intelligence for UAV-Enabled Wireless Networks: A Survey
2021
IEEE Open Journal of the Communications Society
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2021; you can also visit the original URL.
The file type is application/pdf.
In this article, we provide a comprehensive overview of some potential applications of AI in UAV-based networks. ...
As a result, a significant part of the research community has started to integrate intelligence at the core of UAVs networks by applying AI algorithms in solving several problems in relation to drones. ...
transforming data from a high-dimensional space representation to a lower-dimensional space. ...
doi:10.1109/ojcoms.2021.3075201
fatcat:4q6cl2sz7nha5mijm6fmomv3fi
Acoustic Features for Environmental Sound Analysis
[chapter]
2017
Computational Analysis of Sound Scenes and Events
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2018; you can also visit the original URL.
The file type is application/pdf.
Therefore feature selection is generally solved in a sub-optimal manner, usually by introducing two main simplifications: • Brute-force search is avoided by recurring to a near-optimal search strategy. ...
Dimensionality reduction A common approach to cope with the potentially large dimensionality of the feature space is to use transformation techniques such as PCA, linear discriminant analysis (LDA), or ...
doi:10.1007/978-3-319-63450-0_4
fatcat:mudkectf6nbp7hdab7u75kv7ei
A Journey from Improper Gaussian Signaling to Asymmetric Signaling
2020
IEEE Communications Surveys and Tutorials
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
Note that this fulltext copy is not of the "primary" version of this work. The version it corresponds to is:
2020 pre-print arXiv:2005.00058v1 fatcat:mxsuueh3vjeizmmd75n5z7byku
As such, the theory of impropriety has vast applications in medicine, geology, acoustics, optics, image and pattern recognition, computer vision, and other numerous research fields with our main focus ...
The deviation of continuous and discrete complex random variables from the traditional proper and symmetric assumption to a generalized improper and asymmetric characterization (accounting correlation ...
For minimal dimensions, line search or even exhaustive search can give promising results whereas other algorithms are needed for NP-hard problems. ...
doi:10.1109/comst.2020.2989626
fatcat:zyno7ku6n5eqnp6rrcopczb4qu
Next Generation Terahertz Communications: A Rendezvous of Sensing, Imaging, and Localization
2020
IEEE Communications Magazine
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
In this paper, we present a progressive vision of how the traditional "THz gap" will transform into a "THz rush" over the next few years. ...
Then, we illustrate how their coalescence results in enhanced environment-aware system performance in beyond-5G use cases. ...
Based on the pairwise estimated distances, the MDS algorithm tries to locate the sensor nodes in a given dimensional space. ...
doi:10.1109/mcom.001.1900698
fatcat:dhllcybc6fcpbkowo7j6dtz6ey
Next Generation Terahertz Communications: A Rendezvous of Sensing, Imaging, and Localization
[article]
2020
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
Other Versions
In this paper, we present a progressive vision of how the traditional "THz gap" will transform into a "THz rush" over the next few years. ...
Then, we illustrate how their coalescence results in enhanced environment-aware system performance in beyond-5G use cases. ...
Based on the pairwise estimated distances, the MDS algorithm tries to locate the sensor nodes in a given dimensional space. ...
arXiv:1909.10462v2
fatcat:aub5wqfibndtnhz54meg45c7ya
Smart Radio Environments Empowered by Reconfigurable Intelligent Surfaces: How it Works, State of Research, and Road Ahead
[article]
2020
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
What are the most suitable uses and applications of reconfigurable intelligent surfaces in wireless networks? What are the most promising smart radio environments for wireless applications? ...
Maxwell's mathematical theories of electromagnetism, and reporting pragmatic guidelines and recipes for employing appropriate physics-based models of metasurfaces in wireless communications. ...
The general case study that encompasses two-dimensional metasurfaces in a three-dimensional space can be found in [43] . ...
arXiv:2004.09352v1
fatcat:ncsxqispbrcdrcpyrqvja3ie3m
Multiview Approaches to Event Detection and Scene Analysis
[chapter]
2017
Computational Analysis of Sound Scenes and Events
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2018; you can also visit the original URL.
The file type is application/pdf.
Joint subspace learning
Feature-space transformation A number of techniques has been suggested to map the observed feature vectors from two modalities to a low dimensional space where a measure of "dependency ...
the same space, dimensionality difference between views is eliminated and direct comparison across views is made possible. ...
doi:10.1007/978-3-319-63450-0_9
fatcat:3s3dchsicbg4pkuqrsnrvtkrs4
A Key 6G Challenge and Opportunity – Connecting the Remaining 4 Billions: A Survey on Rural Connectivity
[article]
2019
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2019; you can also visit the original URL.
The file type is application/pdf.
In addition, energy and cost efficiency of the studied technologies are analyzed. Typical application scenarios in rural areas are discussed, and several country-specific use cases are surveyed. ...
Providing connectivity to around half of the World population living in rural or underprivileged areas is a tremendous challenge, but also a unique opportunity. ...
Several optimization algorithms (Hill Climbing, Virtual Force, Time-Efficient Local Search, and Random) were implemented for different rural settlement models (Dispersed, Linear, Nucleated, and Isolated ...
arXiv:1906.11541v1
fatcat:kb6tpkvskjggbp25bt7it5iibm
« Previous
Showing results 1 — 15 out of 2,311 results