WO2024135861A1 - Procédé d'entraînement de réseau d'apprentissage profond appliquant un type de représentation de données variable, et dispositif mobile l'appliquant - Google Patents

Procédé d'entraînement de réseau d'apprentissage profond appliquant un type de représentation de données variable, et dispositif mobile l'appliquant Download PDF

Info

Publication number
WO2024135861A1
WO2024135861A1 PCT/KR2022/020665 KR2022020665W WO2024135861A1 WO 2024135861 A1 WO2024135861 A1 WO 2024135861A1 KR 2022020665 W KR2022020665 W KR 2022020665W WO 2024135861 A1 WO2024135861 A1 WO 2024135861A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
deep learning
phenotype
layer
learning network
Prior art date
Application number
PCT/KR2022/020665
Other languages
English (en)
Korean (ko)
Inventor
이상설
장성준
박종희
Original Assignee
한국전자기술연구원
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 한국전자기술연구원 filed Critical 한국전자기술연구원
Priority claimed from KR1020220177737A external-priority patent/KR20240096949A/ko
Publication of WO2024135861A1 publication Critical patent/WO2024135861A1/fr

Links

Images

Definitions

  • the present invention relates to deep learning learning, and more specifically, to a method of additionally learning and re-learning a model that has been trained on a server on a mobile device.
  • re-learning In order to operate a model that has been trained on a server on a new device, re-learning must be performed to regenerate deep learning parameters using the data used for learning and the data used for testing.
  • the present invention was created to solve the above problems, and the purpose of the present invention is to provide a method for additional learning or relearning in mobile devices with limited memory resources, and a deep learning computing device and method using variable data phenotypes. In providing.
  • a deep learning network learning method includes the steps of setting a data phenotype for each layer constituting a deep learning network; It includes the step of training a deep learning network while changing the phenotype of the data input to each layer according to the set data phenotype.
  • the setting step may be setting the number of bits in the exponent part and setting the number of bits in the mantissa part for each layer.
  • the setting step may be setting the data representation for each channel for each layer.
  • the deep learning network training method according to the present invention may further include quantizing a training dataset to be used for deep learning network training.
  • the learning step may be to modify the phenotype of the data between layers from the phenotype of the previous layer to the phenotype of the next layer.
  • the setting step may be setting the data expression for each layer based on the data expression received from the outside.
  • Data phenotype can be determined based on the importance of each layer.
  • a control unit that sets data phenotypes for each layer constituting a deep learning network; and a data conversion unit that changes the phenotype of data input to each layer according to the data phenotype set by the control unit.
  • a deep learning network device comprising a deep learning accelerator that trains a deep learning network with data whose phenotype is changed by a data conversion unit.
  • variable data phenotypes for each layer and channel when performing additional learning or relearning, there is a significant reduction in accuracy in mobile devices with limited memory resources and power. It is possible to perform additional learning and retraining of the deep learning model without it.
  • 1 is a diagram showing the inference process (forward path) in the learning process
  • Figure 2 is a diagram showing the backpropagation process (Backward path) in the learning process
  • Figure 6 shows learning performance using data suggested phenotypes
  • Figure 7 is a mobile device according to an embodiment of the present invention.
  • FP16 which usually uses fewer bits, is used, or hardware is configured using fixed points to reduce the size of the operator.
  • this also has the disadvantage of slowing the learning speed due to bandwidth problems occurring when inputting and outputting data to and from external memory, and lowering the learning accuracy due to the low accuracy of the calculator.
  • an embodiment of the present invention presents a deep learning learning method applying variable data phenotypes.
  • This is a technology that processes data (storage, calculation, transmission, etc.) by changing it into a flexible data format according to a specific distribution so that various phenotypes (fixed point, floating point, etc.) shown in Figure 4 can be applied for learning data processing and calculation.
  • the data expression can be set differently for each layer. By distinguishing between important and non-important layers, the data expression can be set differently.
  • data expression is set for each channel.
  • the data phenotype can be set differently for each channel. By distinguishing between important channels and unimportant channels, the data phenotype can be set differently.
  • Figure 5 shows variable data phenotypes that can be set.
  • PEE Partial Exponent Expression
  • Revised FP represents the mantissa part
  • the data expression can be set by flexibly changing the number of bits in both the exponent part and the mantissa part.
  • the data phenotype can be set/controlled by receiving the data phenotype set from an external source, such as the host.
  • Revised FP size of Revised FP can be changed to a variety of bits depending on hardware resources.
  • Revised FP if the operation is performed with an existing fixed point, all bits are considered as mantissa, and when processing with FP operation, small exponent expression and mantissa are used. It can be expressed by dividing it into .
  • the deep learning network is trained by changing the phenotype of the data input to each layer according to the set data phenotype, that is, changing the phenotype of the data between layers from the phenotype of the previous layer to the phenotype of the next layer.
  • the method according to the embodiment of the present invention is mainly used in the low-bit or data pre-processing process during hardware processing, it can operate similarly to existing data input by adding simple hardware logic, resulting in performance similar to original-level learning. It is possible to derive
  • memory usage can be further reduced by using the minimum number of bits in the learning process, and the intermediate data storage space for the learning data can be used as additional batch data. It is also possible to increase the size.
  • FIG. 7 is a block diagram of a mobile device according to an embodiment of the present invention.
  • a mobile device according to an embodiment of the present invention is configured to include a memory 110, an MCU 120, a deep learning accelerator 130, a quantization unit 140, and a data conversion unit 150, as shown. .
  • the MCU 120 receives some learning data from the server, stores it in the memory 110, and sets the data phenotype for each layer/channel of the deep learning network.
  • the quantization unit 140 performs quantization on the training data stored in the memory 110, and the data conversion unit 150 converts the quantized training data into a phenotype set in each layer/channel.
  • the deep learning accelerator 130 further trains or retrains the deep learning model learned on the server using learning data that is quantized by the quantization unit 140 and then converted into phenotype by the data conversion unit 150.
  • variable data phenotypes are used for each layer/channel. It was made applicable.
  • a computer-readable recording medium can be any data storage device that can be read by a computer and store data.
  • computer-readable recording media can be ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical disk, hard disk drive, etc.
  • computer-readable codes or programs stored on a computer-readable recording medium may be transmitted through a network connected between computers.

Landscapes

  • Mobile Radio Communication Systems (AREA)

Abstract

Sont proposés un procédé d'entraînement de réseau d'pprentissage profond appliquant un type de représentation de données variable, et un dispositif mobile l'appliquant. Le procédé d'entraînement de réseau d'apprentissage profond selon un mode de réalisation de la présente invention consiste à : configurer un type de représentation de données pour chacune de couches constituant un réseau d'apprentissage profond ; et entraîner le réseau d'apprentissage profond tout en changeant un type de représentation de données, qui est entré dans chaque couche, en fonction du type de représentation de données configuré. Par conséquent, en appliquant un type de représentation de données variable pour chaque couche et chaque canal lors de la réalisation d'un apprentissage supplémentaire ou d'un réapprentissage, il est possible de réaliser un apprentissage supplémentaire et un réapprentissage pour un modèle d'apprentissage profond sur un dispositif mobile avec des ressources de mémoire et une puissance limitées sans réduction significative de la précision.
PCT/KR2022/020665 2022-12-19 2022-12-19 Procédé d'entraînement de réseau d'apprentissage profond appliquant un type de représentation de données variable, et dispositif mobile l'appliquant WO2024135861A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
KR1020220177737A KR20240096949A (ko) 2022-12-19 가변 데이터 표현형을 적용한 딥러닝 학습 방법 및 이를 적용한 모바일 디바이스
KR10-2022-0177737 2022-12-19

Publications (1)

Publication Number Publication Date
WO2024135861A1 true WO2024135861A1 (fr) 2024-06-27

Family

ID=91589081

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/KR2022/020665 WO2024135861A1 (fr) 2022-12-19 2022-12-19 Procédé d'entraînement de réseau d'apprentissage profond appliquant un type de représentation de données variable, et dispositif mobile l'appliquant

Country Status (1)

Country Link
WO (1) WO2024135861A1 (fr)

Similar Documents

Publication Publication Date Title
CN110969251B (zh) 基于无标签数据的神经网络模型量化方法及装置
JP2020528625A (ja) 翻訳方法、ターゲット情報決定方法および関連装置、記憶媒体
CN111030861B (zh) 一种边缘计算分布式模型训练方法、终端和网络侧设备
KR102655950B1 (ko) 뉴럴 네트워크의 고속 처리 방법 및 그 방법을 이용한 장치
WO2020159016A1 (fr) Procédé d'optimisation de paramètre de réseau neuronal approprié pour la mise en œuvre sur matériel, procédé de fonctionnement de réseau neuronal et appareil associé
CN108446766A (zh) 一种快速训练堆栈自编码深度神经网络的方法
EP4008057B1 (fr) Compression de poids d'exposant sans perte et de poids de mantisse avec perte d'apprentissage de réseaux neuronaux profonds
WO2022146080A1 (fr) Algorithme et procédé de modification dynamique de la précision de quantification d'un réseau d'apprentissage profond
WO2024135861A1 (fr) Procédé d'entraînement de réseau d'apprentissage profond appliquant un type de représentation de données variable, et dispositif mobile l'appliquant
CN112860904A (zh) 一种融入外部知识的生物医疗关系抽取方法
EP4227850A1 (fr) Programme, procédé d'apprentissage et appareil de traitement d'informations
KR20240096949A (ko) 가변 데이터 표현형을 적용한 딥러닝 학습 방법 및 이를 적용한 모바일 디바이스
WO2023003432A1 (fr) Procédé et dispositif pour déterminer une plage de quantification basée sur un taux de saturation pour la quantification d'un réseau de neurones
WO2023085457A1 (fr) Structure de mémoire et procédé de commande pour formation à apprentissage profond efficace
WO2023277448A1 (fr) Procédé et système d'entraînement de modèle de réseau neuronal artificiel pour traitement d'image
WO2023080292A1 (fr) Appareil et procédé pour générer un paramètre adaptatif pour un dispositif d'accélération d'apprentissage profond
CN110852361B (zh) 基于改进深度神经网络的图像分类方法、装置与电子设备
WO2023085458A1 (fr) Procédé et dispositif de commande d'une mémoire de formation d'apprentissage profond allégée
WO2022107951A1 (fr) Procédé de formation d'un réseau d'apprentissage profond ultra-léger
WO2023128024A1 (fr) Procédé et système de quantification de réseau d'apprentissage profond
WO2022107910A1 (fr) Dispositif matériel mobile d'apprentissage profond apte à un réentraînement
WO2023177026A1 (fr) Procédé et appareil destinés à prédire la performance d'un réseau neuronal artificiel en fonction d'un format de données
WO2024135870A1 (fr) Dispositif de reconnaissance d'image exécutant un procédé de quantification de réseau d'unités d'entrée pour une détection efficace d'objet
CN112257469A (zh) 用于小型移动设备的深层神经机器翻译模型的压缩方法
WO2023014124A1 (fr) Procédé et appareil de quantification d'un paramètre de réseau neuronal