Filters








727 Hits in 6.6 sec

Deep Constrained Siamese Hash Coding Network and Load-Balanced Locality-Sensitive Hashing for Near Duplicate Image Detection

Weiming Hu, Yabo Fan, Junliang Xing, Liang Sun, Zhaoquan Cai, Stephen Maybank
2018 IEEE Transactions on Image Processing  
The extracted features are used to construct a LSH-based index. We propose a load-balanced LSH method to produce load-balanced buckets in the hashing process.  ...  We propose a deep constrained siamese hash coding neural network combined with deep feature learning. Our neural network is able to extract effective features for near duplicate image detection.  ...  This accelerates the detection and also obtains good accuracy. Therefore, our load-balanced LSH is efficient and flexible in contrast with the basic LSH.  ... 
doi:10.1109/tip.2018.2839886 pmid:29897871 fatcat:2kdkcu53vjbunohy7fvoyvwfv4

Fast GPU-based locality sensitive hashing for k-nearest neighbor computation

Jia Pan, Dinesh Manocha
2011 Proceedings of the 19th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems - GIS '11  
In the query step, we use GPU-based work queues to accelerate short-list search, which is one of the main bottlenecks in LSH-based algorithms.  ...  In practice, our GPU implementation can obtain more than 40X acceleration over a single-core CPU-based LSH implementation.  ...  The final ranking computation among the candidates is called the short-list search, which is regarded as the main bottleneck in LSH-based algorithms.  ... 
doi:10.1145/2093973.2094002 dblp:conf/gis/PanM11 fatcat:s4cm3nz7qjerrkhjdou7iipyr4

A Survey on Efficient Processing of Similarity Queries over Neural Embeddings [article]

Yifan Wang
2022 arXiv   pre-print
Finally, we investigate the specific solutions with and without using embeddings in selected application domains of similarity queries, including entity resolution and information retrieval.  ...  semantics of the raw data, based on which embeddings do show outstanding effectiveness on capturing data similarities, making it one of the most widely used and studied techniques in the state-of-the-art  ...  With the hashing based index emerging in high-dimensional KNN search, it is also used in KNN join approaches. RankReduce [187] implements a distributed LSH based KNN search system using MapReduce.  ... 
arXiv:2204.07922v1 fatcat:u5osyghs6vgppnj5gpnrzhae5y

A Tale of Two Efficient and Informative Negative Sampling Distributions [article]

Shabnam Daghaghi, Tharun Medini, Nicholas Meisburger, Beidi Chen, Mengnan Zhao, Anshumali Shrivastava
2021 arXiv   pre-print
Softmax classifiers with a very large number of classes naturally occur in many applications such as natural language processing and information retrieval.  ...  In this paper, we show two classes of distributions where the sampling scheme is truly adaptive and provably generates negative samples in near-constant time.  ...  Definition 1 Adaptive Negative Sampling: We call a negative sampling distribution adaptive if the distribution changes with the change in the parameter of the network as well as the change in the input  ... 
arXiv:2012.15843v2 fatcat:6bbc72stgbezhohwiogoqpxcoe

Scalable Locality-Sensitive Hashing for Similarity Search in High-Dimensional, Large-Scale Multimedia Datasets [article]

Thiago S. F. X. Teixeira, George Teodoro, Eduardo Valle, Joel H. Saltz
2013 arXiv   pre-print
Here we address these issues with a distributed, efficient, and scalable index based on Locality-Sensitive Hashing (LSH).  ...  Thus, scalability is imperative for similarity search in Web-scale applications, but most existing methods are sequential and target shared-memory machines.  ...  Similarity search is a core operation for content-based multimedia retrieval (CBMR) applications such as image search engines, realtime song identification, tagging of photos in social networks, etc.  ... 
arXiv:1310.4136v1 fatcat:xb5vsm4dg5estlyhlkyievhdl4

Entropy based locality sensitive hashing

Qiang Wang, Zhiyuan Guo, Gang Liu, Jun Guo
2012 2012 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)  
Locality sensitive hashing (LSH) scheme based on p-stable distributions is a good solution to the approximate nearest neighbor (ANN) problem, but points are always mapped to a poor distribution.  ...  This paper proposes a set of new hash mapping functions based on entropy for LSH.  ...  Indy presented LSH scheme based on p-stable distributions in original non-hamming space [3] .  ... 
doi:10.1109/icassp.2012.6288065 dblp:conf/icassp/WangGLG12 fatcat:mqpwqtun3nbl7idbjit6rj2caq

Benchmark of DNN Model Search at Deployment Time [article]

Lixi Zhou, Arindam Jain, Zijie Wang, Amitabh Das, Yingzhen Yang, Jia Zou
2022 arXiv   pre-print
This paper proposes multiple model search strategies including various similarity-based approaches and non-similarity-based approaches.  ...  The experimental evaluation showed that our proposed asymmetric similarity-based measurement, adaptivity, outperformed symmetric similarity-based measurements and non-similarity-based measurements in most  ...  We further propose a novel two-level LSH index framework to accelerate the system-wide computation of adaptivity with theoretical proofs.  ... 
arXiv:2206.00188v1 fatcat:3tawlkzhwzebhbojr76f6uuuze

Hardware-Software Co-Design of an In-Memory Transformer Network Accelerator

Ann Franchesca Laguna, Mohammed Mehdi Sharifi, Arman Kazemi, Xunzhao Yin, Michael Niemier, X. Sharon Hu
2022 Frontiers in Electronics  
We propose an in-memory transformer network accelerator (iMTransformer) that uses a combination of crossbars and content-addressable memories to accelerate transformer networks.  ...  We accelerate transformer networks by (1) computing in-memory, thus minimizing the memory transfer overhead, (2) caching reusable parameters to reduce the number of operations, and (3) exploiting the available  ...  AL wrote the manuscript with input from all authors. All authors reviewed the document.  ... 
doi:10.3389/felec.2022.847069 fatcat:vtlssmfnenbvzk24qyjbrm2bgm

Bootstrap sequential projection multi kernel Locality Sensitive Hashing

Harsham Mehta, Deepak Garg
2014 2014 International Conference on Advances in Computing, Communications and Informatics (ICACCI)  
Hence it approximated and accelerated the traditional search strategy by introducing the hamming space computation. It is liable to consume time complexity in modern CPU architecture.  ...  Both hash table construction and searching in it are distributed across multiple cores and multiple nodes in PLSH. 2.  ... 
doi:10.1109/icacci.2014.6968294 dblp:conf/icacci/MehtaG14 fatcat:hjnzm5fayfesjmaaau33w2bikq

Experimentally realized memristive memory augmented neural network [article]

Ruibin Mao
2022 arXiv   pre-print
Previous works on emerging memory-based implementation have difficulties in scaling up because different modules with various structures are difficult to integrate on the same chip and the small sense  ...  Memory augmented neural network has been proposed to achieve the goal, but the memory module has to be stored in an off-chip memory due to its size.  ...  We demonstrate that analog in-memory computing with memristive crossbars efficiently supports many different tasks, including convolution, hashing, and content-based searching.  ... 
arXiv:2204.07429v1 fatcat:5xvdijrhb5bple54gvcpnapmum

Locality-sensitive hashing in function spaces [article]

Will Shand, Stephen Becker
2020 arXiv   pre-print
We discuss the problem of performing similarity search over function spaces. To perform search over such spaces in a reasonable amount of time, we use locality-sensitive hashing (LSH).  ...  We use the presented hashing schemes to construct an LSH family for Wasserstein distance over one-dimensional, continuous probability distributions.  ...  LSH offers a promising method to accelerate similarity search with Wasserstein distance.  ... 
arXiv:2002.03909v1 fatcat:6jave56byzaidasogrlkboz5ja

Similarity Search with Tensor Core Units [article]

Thomas D. Ahle, Francesco Silvestri
2020 arXiv   pre-print
In this paper, we show that TCUs can speed up similarity search problems as well.  ...  Tensor Core Units (TCUs) are hardware accelerators developed for deep neural networks, which efficiently support the multiplication of two dense √(m)×√(m) matrices, where m is a given hardware parameter  ...  When τ = O(m), the TCU algorithm exhibits a √ m speedup with respect to traditional approaches (even those based on LSH).  ... 
arXiv:2006.12608v1 fatcat:qji5hmb7wvgnjdiy2l3autqh4e

It's the Best Only When It Fits You Most: Finding Related Models for Serving Based on Dynamic Locality Sensitive Hashing [article]

Lixi Zhou, Zijie Wang, Amitabh Das, Jia Zou
2020 arXiv   pre-print
We find that our proposed adaptivity measurement which is based on Jensen-Shannon (JS) divergence, is an effective measurement, and its computation can be significantly accelerated by using the technique  ...  They are in lack of an automatic model searching tool.  ...  Conclusions In this work, we systematically explore the problem of finding related models for serving based on JS-divergence and adaptivity, which are dynamically computed over the features shared by the  ... 
arXiv:2010.09474v1 fatcat:d7x7zsfcdfhpppja6sefcjjbli

MONGOOSE: A Learnable LSH Framework for Efficient Neural Network Training

Beidi Chen, Zichang Liu, Binghui Peng, Zhaozhuo Xu, Jonathan Lingjie Li, Tri Dao, Zhao Song, Anshumali Shrivastava, Christopher Ré
2021 International Conference on Learning Representations  
However, while LSH has sublinear guarantees for approximate near-neighbor search in theory, it is known to have inefficient query time in practice due to its use of random hash functions.  ...  Recent advances by practitioners in the deep learning community have breathed new life into Locality Sensitive Hashing (LSH), using it to reduce memory and time bottlenecks in neural network (NN) training  ...  Government is authorized to reproduce and distribute reprints for Governmental purposes notwithstanding any copyright notation thereon.  ... 
dblp:conf/iclr/ChenLPXLD0SR21 fatcat:j2cuvjt66jbtxmxjux65rkhyfy

FLASH: Randomized Algorithms Accelerated over CPU-GPU for Ultra-High Dimensional Similarity Search [article]

Yiqiu Wang, Anshumali Shrivastava, Jonathan Wang, Junghee Ryu
2018 arXiv   pre-print
We present FLASH (Fast LSH Algorithm for Similarity search accelerated with HPC), a similarity search system for ultra-high dimensional datasets on a single machine, that does not require similarity computations  ...  By leveraging a LSH style randomized indexing procedure and combining it with several principled techniques, such as reservoir sampling, recent advances in one-pass minwise hashing, and count based estimations  ...  We would like to thank anonymous reviewers and Rasmus Pagh for discussions on the role of correlations in Theorem 1.  ... 
arXiv:1709.01190v2 fatcat:vic3h5gjnbfnppyruoc4io6rsu
« Previous Showing results 1 — 15 out of 727 results