BR112023019150A2 - Técnica de compressão para ponderações de rede neural profunda - Google Patents

Técnica de compressão para ponderações de rede neural profunda

Info

Publication number
BR112023019150A2
BR112023019150A2 BR112023019150A BR112023019150A BR112023019150A2 BR 112023019150 A2 BR112023019150 A2 BR 112023019150A2 BR 112023019150 A BR112023019150 A BR 112023019150A BR 112023019150 A BR112023019150 A BR 112023019150A BR 112023019150 A2 BR112023019150 A2 BR 112023019150A2
Authority
BR
Brazil
Prior art keywords
weighting
frame
compressed
data
weighting data
Prior art date
Application number
BR112023019150A
Other languages
English (en)
Inventor
Haoping Xu
Narayana Macha Lakshmi
Prajakt Kulkarni
Original Assignee
Qualcomm Inc
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Qualcomm Inc filed Critical Qualcomm Inc
Publication of BR112023019150A2 publication Critical patent/BR112023019150A2/pt

Links

Classifications

    • H—ELECTRICITY
    • H03—ELECTRONIC CIRCUITRY
    • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M7/00—Conversion of a code where information is represented by a given sequence or number of digits to a code where the same, similar or subset of information is represented by a different sequence or number of digits
    • H03M7/30—Compression; Expansion; Suppression of unnecessary data, e.g. redundancy reduction
    • H03M7/70—Type of the data to be coded, other than image and sound
    • H03M7/702—Software
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0495—Quantised networks; Sparse networks; Compressed networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
    • G06N3/063—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using electronic means
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • H—ELECTRICITY
    • H03—ELECTRONIC CIRCUITRY
    • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M7/00—Conversion of a code where information is represented by a given sequence or number of digits to a code where the same, similar or subset of information is represented by a different sequence or number of digits
    • H03M7/30—Compression; Expansion; Suppression of unnecessary data, e.g. redundancy reduction
    • H03M7/46—Conversion to or from run-length codes, i.e. by representing the number of consecutive digits, or groups of digits, of the same kind by a code word and a digit indicative of that kind

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • General Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Computational Linguistics (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Neurology (AREA)
  • Compression, Expansion, Code Conversion, And Decoders (AREA)

Abstract

técnica de compressão para ponderações de rede neural profunda. várias modalidades incluem métodos e dispositivos para compressão e descompressão de conjuntos de dados de ponderação. algumas modalidades podem incluir compressão de dados de ponderação recebendo um conjunto de dados de ponderação de números binários que representam valores de ponderação, gerar uma carga útil de quadro incluindo um primeiro quadro comprimido de um primeiro subconjunto dos valores de ponderação no conjunto de dados de ponderação, e gerar um bloco de dados de ponderação comprimidos tendo a carga útil de quadro. algumas modalidades podem incluir descompressão de dados de ponderação recuperando um bloco de dados de ponderação comprimidos, em que o bloco de dados de ponderação comprimidos inclui um cabeçalho de quadro associado a uma carga útil de quadro, em que o cabeçalho de quadro inclui um indicador de fator de normalização, e em que a carga útil de quadro inclui valores de ponderação comprimidos, e gerar um primeiro quadro descomprimido compreendendo valores de ponderação descomprimidos dos valores de ponderação comprimidos da carga útil de quadro.
BR112023019150A 2021-04-01 2022-03-30 Técnica de compressão para ponderações de rede neural profunda BR112023019150A2 (pt)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US17/220,620 US11757469B2 (en) 2021-04-01 2021-04-01 Compression technique for deep neural network weights
PCT/US2022/022497 WO2022212467A1 (en) 2021-04-01 2022-03-30 Compression technique for deep neural network weights

Publications (1)

Publication Number Publication Date
BR112023019150A2 true BR112023019150A2 (pt) 2023-10-17

Family

ID=81308565

Family Applications (1)

Application Number Title Priority Date Filing Date
BR112023019150A BR112023019150A2 (pt) 2021-04-01 2022-03-30 Técnica de compressão para ponderações de rede neural profunda

Country Status (7)

Country Link
US (1) US11757469B2 (pt)
EP (1) EP4315175A1 (pt)
JP (1) JP2024514448A (pt)
KR (1) KR20230162778A (pt)
CN (1) CN117099109A (pt)
BR (1) BR112023019150A2 (pt)
WO (1) WO2022212467A1 (pt)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US12224774B2 (en) * 2022-11-16 2025-02-11 Samsung Electronics Co., Ltd. Runtime reconfigurable compression format conversion
US20240223785A1 (en) * 2022-12-29 2024-07-04 Samsung Electronics Co., Ltd. Electronic device and method with image encoding and decoding

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101498063B1 (ko) * 2008-03-04 2015-03-03 엘지전자 주식회사 디지털 방송 시스템 및 데이터 처리 방법
US8489961B2 (en) * 2009-10-19 2013-07-16 Lg Electronics Inc. Transmitting system and method of processing digital broadcast signal in transmitting system, receiving system and method of receiving digital broadcast signal in receiving system
WO2012003602A1 (zh) * 2010-07-09 2012-01-12 西安交通大学 一种电子喉语音重建方法及其系统
US10608664B2 (en) 2018-05-09 2020-03-31 Samsung Electronics Co., Ltd. Electronic apparatus for compression and decompression of data and compression method thereof
US11625245B2 (en) * 2018-09-28 2023-04-11 Intel Corporation Compute-in-memory systems and methods
CN111178490B (zh) * 2019-12-31 2021-08-24 北京百度网讯科技有限公司 数据输出方法、获取方法、装置和电子设备
US11925128B2 (en) * 2020-08-26 2024-03-05 Robert Bosch Gmbh Differential ionic electronic transistors
US12314859B2 (en) * 2020-10-08 2025-05-27 Samsung Electronics Co., Ltd. Electronic apparatus for decompressing a compressed artificial intelligence model and control method therefor

Also Published As

Publication number Publication date
JP2024514448A (ja) 2024-04-02
WO2022212467A1 (en) 2022-10-06
KR20230162778A (ko) 2023-11-28
US11757469B2 (en) 2023-09-12
US20220321143A1 (en) 2022-10-06
EP4315175A1 (en) 2024-02-07
CN117099109A (zh) 2023-11-21

Similar Documents

Publication Publication Date Title
BR112023019150A2 (pt) Técnica de compressão para ponderações de rede neural profunda
EP3716158A3 (en) Compressing weight updates for decoder-side neural networks
WO2006056974A3 (en) Xml parser
PH12019550191A1 (en) Neural network processor using compression and decompression of activation data to reduce memory bandwidth utilization
US20100106269A1 (en) Method and apparatus for signal processing using transform-domain log-companding
EP2424117A2 (en) Sparse data compression
EP1628290A3 (en) Generation of a filterbank for audio compression
IN2014DE02379A (pt)
MY204413A (en) Method and apparatus for storing data, and computer device and storage medium thereof
US10978091B2 (en) System and methods for suppression by selecting wavelets for feature compression in distributed speech recognition
EP4280604A3 (en) Guaranteed data compression
Jiang et al. Speech slytherin: Examining the performance and efficiency of mamba for speech separation, recognition, and synthesis
US10417287B2 (en) Compressing short text messages
Shukla et al. Audio compression algorithm using discrete cosine transform (DCT) and Lempel-Ziv-Welch (LZW) encoding method
CN113178191B (zh) 基于联邦学习的语音表征模型训练方法、装置、设备及介质
Ferragina et al. The engineering of a compression boosting library: Theory vs practice in BWT compression
EP1768104A1 (en) Signal encoding device and method, and signal decoding device and method
Aji et al. Neural machine translation with 4-bit precision and beyond
Beirami et al. Memory-assisted universal source coding
WO2023030557A3 (zh) 数据压缩、解压缩方法及装置
Perić et al. Design of fixed and adaptive companding quantizer with variable-length codeword for memoryless Gaussian source
Firouzeh et al. Improved recovery of compressive sensed speech
Axelsson et al. File fragment analysis using normalized compression distance
Kumar et al. A compressive sensing codec architecture for ECG signals with adaptive quantization and stream entropy coding
Reynolds Model compression for GMM based speaker recognition systems.