A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2017; you can also visit the original URL.
The file type is application/pdf.
Filters
Additive Quantization for Extreme Vector Compression
2014
2014 IEEE Conference on Computer Vision and Pattern Recognition
Preserved Fulltext
We introduce a new compression scheme for highdimensional vectors that approximates the vectors using sums of M codewords coming from M different codebooks. ...
In the experiments, we demonstrate that the proposed compression can be used instead of or together with product quantization. ...
Additive product quantization The complexity of the Beam Search algorithm grows cubically with M . While for the extreme compression (e.g. ...
doi:10.1109/cvpr.2014.124
dblp:conf/cvpr/BabenkoL14
fatcat:x4azagnsmnhevlcodyojn54euq
Training with Quantization Noise for Extreme Model Compression
[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
Controlling the amount of noise and its form allows for extreme compression rates while maintaining the performance of the original model. ...
In this paper, we extend this approach to work beyond int8 fixed-point quantization with extreme compression methods where the approximations introduced by STE are severe, such as Product Quantization. ...
For language modeling, we train for 10 additional epochs. For RoBERTa, we train for 25k additional updates. ...
arXiv:2004.07320v3
fatcat:lexa56jkmfffvgnyl3f7dfonkq
Extremely Low Bit-Rate Nearest Neighbor Search Using a Set Compression Tree
2014
IEEE Transactions on Pattern Analysis and Machine Intelligence
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 2013; you can also visit the original URL.
The file type is application/pdf.
, and Iterative Quantization. ...
The goal of this work is a data structure to support approximate nearest neighbor search on very large scale sets of vector descriptors. ...
We are grateful for financial support from ERC grant VisRec no. 228180. ...
doi:10.1109/tpami.2014.2339821
pmid:26353147
fatcat:qnt6vtbt35dddohgoryora7h6e
CosSGD: Communication-Efficient Federated Learning with a Simple Cosine-Based Quantization
[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
Further, our approach is highly suitable for federated learning problems since it has low computational complexity and requires only a little additional data to recover the compressed information. ...
when quantization are applied in double directions to compress model weights and gradients. ...
In addition to setting the exact b θ for a vector, we also clip the top dimensions alternatively. Sometimes, there is one dimension dominating the gradient or weight vector. ...
arXiv:2012.08241v2
fatcat:pes6bfrkxveohcouipxa4uwdga
Compressing Deep Convolutional Networks using Vector Quantization
[article]
2014
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.
In this paper, we tackle this model storage issue by investigating information theoretical vector quantization methods for compressing the parameters of CNNs. ...
In particular, we have found in terms of compressing the most storage demanding dense connected layers, vector quantization methods have a clear gain over existing matrix factorization methods. ...
RQ works extremely poorly for such a task, which probably means there are few global structures in these weight vectors. ...
arXiv:1412.6115v1
fatcat:qmfcwljfjjaubmfjw3mxgegn2y
Feature Vector Compression Based on Least Error Quantization
2016
2016 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)
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 2016; you can also visit the original URL.
The file type is application/pdf.
We propose a distinctive feature vector compression method based on least error quantization. ...
In this paper, we prove that minimizing quantization error between the compressed and original vectors is most effective to control the performance in face recognition. ...
Therefore, instead of a common quantizer, we efficiently compress each vector with the non-uniform quantizer optimized for the vector, although we have to add another table to each quantized vector. ...
doi:10.1109/cvprw.2016.18
dblp:conf/cvpr/KawaharaY16
fatcat:5ox2aeusp5gyrfduhlyk2kvzxq
SDR: Efficient Neural Re-ranking using Succinct Document Representation
[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.
better compression rates for the same ranking quality. ...
After this token encoding step, we further reduce the size of entire document representations using a modern quantization technique. ...
Depending on the use case, such tradeoffs are highly desirable, allowing for extreme compression rates that minimize the costs of deploying Q&A systems. ...
arXiv:2110.02065v1
fatcat:objlfg2zlbgdphmnghimoxfyj4
Survey Paper on Fractal Image Compression using Block Truncation Coding Technique
2018
International Journal of Computer Applications
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.
BTC algorithm as well as vector quantization method for purpose of multi-leveltechnique for gray and color image. ...
Designing an efficient compression scheme is more critical with the recent growth of computer applications.Modern applications, in addition to high compression ratio, also demand for efficient encoding ...
[3] , Block truncation committal to writing (BTC) has been thought of extremely economical compression technique for many years. ...
doi:10.5120/ijca2018917486
fatcat:2tjh73dqnnavhefnstskfmzwni
Compression of SAR images using KLT, VQ and mixture of principal components
1997
IEE Proceedings - Radar Sonar and Navigation
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 2012; you can also visit the original URL.
The file type is application/pdf.
Two common methods for compressing images are linear block transform coding such as the Karhunen-Loève transform (KLT) and vector quantization (VQ). ...
Like vector quantization, it partitions the input space into a number of non-overlapping regions, while each region is represented by a number of basis vectors in the manner of transform coding. ...
Vector Quantization At the other extreme, VQ is a purely discrete representation of the data. ...
doi:10.1049/ip-rsn:19971175
fatcat:ganmocvyo5h23kxe2xu62epbhq
A SYSTEMATIC IMAGE COMPRESSION IN THE COMBINATION OF LINEAR VECTOR QUANTISATION AND DISCRETE WAVELET TRANSFORM
2014
International Journal of Research in Engineering and Technology
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 2015; you can also visit the original URL.
The file type is application/pdf.
Vector quantisation (VQ) is a novel technique for image compression. VQ is a lossy compression scheme, used to compress image both in spatial domain & frequency domain. ...
The proposed algorithm uses the most effective and simple methods like self organizing maps and linear vector quantization together with the discrete wavelet transform in order to reduce the loss of information ...
Their reciprocal can be implemented using only integer addition and bit shifts, which are extremely fast operation. ...
doi:10.15623/ijret.2014.0304044
fatcat:js377di43jeargbpmopx24t4li
Combined compression and denoising of images using vector quantization
1998
Applications of Digital Image Processing XXI
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 2012; you can also visit the original URL.
The file type is application/pdf.
optimal quantizer for the estimate. ...
What we present in this paper is a simple but sub-optimal vector quantization (VQ) strategy that combines estimation and compression in one efficient step. ...
This paper introduces a joint compression and denoising technique based on non-linear interpolative vector quantization (NLIVQ).' ...
doi:10.1117/12.323206
fatcat:fsjfdiqejbfy7ofd3bemd3oaxq
Bit-level Optimized Neural Network for Multi-antenna Channel Quantization
[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 2020; you can also visit the original URL.
The file type is application/pdf.
extraction and recovery from the perspective of bit-level quantization performance. ...
Quantized channel state information (CSI) plays a critical role in precoding design which helps reap the merits of multiple-input multiple-output (MIMO) technology. ...
CsiNet-Q(L) denotes the CsiNet version which is trained without considering the quantization. For testing CsiNet-Q(L), each entry of the compressed CSI vector is quantized to L bits. ...
arXiv:1909.10730v1
fatcat:rss7xwx7kjc7tkwdaxpbpueeji
Lossy parametric motion model compression for global motion temporal filtering
2012
2012 Picture Coding Symposium
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 2012; you can also visit the original URL.
The file type is application/pdf.
For that, we propose a compression scheme for perspective motion models using transformation before quantization and temporal redundancy reduction and integrate this scheme into a video coding environment ...
A critical issue is the transmission of accurate higher-order motion parameters with as little additional bits as possible to maximize the compression gain of the whole system. ...
This compression scheme transforms a given model to global motion vectors at the corners of each frame. Subsequently, these vectors are quantized. ...
doi:10.1109/pcs.2012.6213354
dblp:conf/pcs/TokKGS12
fatcat:o2r644wdyjgzxmtwxv6oovbfda
Position-based Scaled Gradient for Model Quantization and Pruning
[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
Second, we empirically show that PSG acting as a regularizer to a weight vector is favorable for model compression domains such as quantization and pruning. ...
The experimental results on CIFAR-10/100 and ImageNet datasets show the effectiveness of the proposed PSG in both domains of pruning and quantization even for extremely low bits. ...
• We apply PSG in quantization and pruning and verify the effectiveness of PSG on CIFAR and ImageNet datasets. We also show that PSGD is very effective for extremely low bit quantization. ...
arXiv:2005.11035v4
fatcat:ftahgiooqnfcbe34jmqnzppc4u
An Efficient Coding Method for Teleconferencing Video and Confocal Microscopic Image Sequences
2008
Journal of Computing and Information Technology
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.
The algorithm uses a 3D vector quantization pyramidal codebook-based model with adaptive pyramidal codebook for compression. ...
The adaptive vector quantization algorithm is used to train the codebook for optimal performance with time. ...
The vector quantization and adaptive codebook procedures are explained in Section 2. 3D vector quantization used for compression and encoding of teleconferencing videos and confocal microscopic image sequences ...
doi:10.2498/cit.1000892
fatcat:wsknuacp4bfbll6igui5flhpze
« Previous
Showing results 1 — 15 out of 17,080 results