WO2023238544A1 - モデル管理装置、モデル管理システム及びモデル管理方法 - Google Patents
モデル管理装置、モデル管理システム及びモデル管理方法 Download PDFInfo
- Publication number
- WO2023238544A1 WO2023238544A1 PCT/JP2023/016453 JP2023016453W WO2023238544A1 WO 2023238544 A1 WO2023238544 A1 WO 2023238544A1 JP 2023016453 W JP2023016453 W JP 2023016453W WO 2023238544 A1 WO2023238544 A1 WO 2023238544A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- training data
- data
- machine learning
- model
- information
- Prior art date
- Legal status (The legal status 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 status listed.)
- Ceased
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/60—Protecting data
- G06F21/62—Protecting access to data via a platform, e.g. using keys or access control rules
Definitions
- the present disclosure relates to a model management device, a model management system, and a model management method.
- machine learning technology Due to advances in machine learning technology such as deep learning, machine learning technology is being widely used in various technical fields.
- machine learning technology it is generally known that the performance of a machine learning model can change depending on training data. It is also known that it is difficult to predict changes.
- one objective of the present disclosure is to provide a technique for managing training data used to train a machine learning model.
- One aspect of the present disclosure includes a data specifying unit that acquires data specifying information, and a first training data unit that is trained using first training data specified by the data specifying information with reference to training data usage information.
- a model identifying unit that identifies a machine learning model of the machine learning model;
- the present invention relates to a model management device, comprising: a processing unit that creates a set and executes processing on the first machine learning model.
- FIG. 1 is a schematic diagram illustrating a model management system according to an embodiment of the present disclosure.
- FIG. 2 is a block diagram showing the hardware configuration of a model management device according to an embodiment of the present disclosure.
- FIG. 3 is a block diagram showing the functional configuration of a model management device according to an embodiment of the present disclosure.
- FIG. 4 is a diagram illustrating detailed information of training data according to an embodiment of the present disclosure.
- FIG. 5 is a diagram illustrating usage information according to an embodiment of the present disclosure.
- FIG. 6 is a flowchart showing model management processing according to an embodiment of the present disclosure.
- FIG. 7 is a schematic diagram illustrating a model management system according to another embodiment of the present disclosure.
- a model management device for managing training data used to train a machine learning model is disclosed.
- a model management device is disclosed.
- Another example is a case where performance deteriorates because the environment at the time of data acquisition is different from the environment at the time of operation. For example, because people's lifestyles and behavior patterns have changed before and after the coronavirus pandemic, training data acquired before the coronavirus pandemic may not be appropriate for systems that will be operated after the pandemic. In this case, training data obtained before the coronavirus pandemic may be inappropriate and may degrade the performance of machine learning models. Therefore, the use of training data obtained before the coronavirus pandemic should be avoided, and machine learning models trained with training datasets containing inappropriate training data should be retrained. .
- a case may be considered in which, for example, it becomes clear after training the machine learning model that the annotations (notes, labels, etc.) created by a certain vendor are of poor quality.
- the training data annotated by the vendor is inappropriate and may degrade the performance of the machine learning model.
- the use of training data annotated by the vendor should be avoided, and machine learning models trained with training datasets containing inappropriate training data should be retrained.
- the quality of the training data used to train the machine learning model is determined, and data that does not meet the criteria, is questionable, or differs from the current environment is not used for training in terms of the performance of the machine learning model. It needs to be done as such.
- training data that is not authorized or licensed for commercial use, or that was previously authorized or licensed but whose authorization or license has since become revoked, may be used to train a commercial machine learning model.
- training data may be used to train a commercial machine learning model.
- machine learning model has been used after training a machine learning model.
- the use of unlicensed or unlicensed training data should be avoided, and machine learning models trained with training datasets containing such training data should be retrained. .
- Another example is a case where training data including age and/or gender is used to train a machine learning model when developing AI (Artificial Intelligence) for recruitment.
- AI Artificial Intelligence
- training data that includes age and/or gender should be avoided, and machine learning models trained with training datasets that include such training data should not use inappropriate data. It is preferable to retrain on the removed dataset.
- a model management device uses the inappropriate training data to train a machine learning model. identify machine learning models that were trained using It's okay.
- a model management system 10 includes a training data database (DB) 20, a usage information database (DB) 30, a terminal 40, and a model management device 100, as shown in FIG.
- the training data DB 20 stores training data. Specifically, the training data DB 20 stores training data used to train one or more machine learning models managed by the model management device 100. For example, the training data DB 20 may store detailed information such as the acquisition location, acquisition time, annotation information, license information, personal information protection, and ethical information of each training data in association with the training data.
- the usage information DB 30 stores usage information of training data.
- the usage information may indicate an association between a machine learning model and training data used to train the machine learning model.
- the usage information DB 30 stores usage information indicating usage status of training data for one or more machine learning models managed by the model management device 100.
- the usage information DB 30 may store identification information of training data used to train the machine learning model in association with each machine learning model managed by the model management device 100.
- the terminal 40 may be operated by a user such as a personal computer (PC), tablet, smartphone, etc., such as an administrator of a machine learning model.
- the terminal 40 is connected to the model management device 100 by wire or wirelessly, and by operating the terminal 40, the user can cause the model management device 100 to perform various processes described below. Further, the terminal 40 may receive data specifying information such as identification information for identifying inappropriate data from the user and provide the received data to the model management device 100.
- the model management device 100 When the model management device 100 acquires data specifying information indicating inappropriate training data etc. from the terminal 40, the model management device 100 refers to the usage information of the training data stored in the usage information DB 30, and selects the training data specified by the data specifying information. Identify machine learning models trained on your data. Then, the model management device 100 updates the training data set by deleting the specified training data from the training data set stored in the training data DB 20 used to train the specified machine learning model. , executes processing on the identified machine learning model (for example, retraining the machine learning model using an updated training data set), and reports the processing results to the terminal 40.
- the model management device 100 When the model management device 100 acquires data specifying information indicating inappropriate training data etc. from the terminal 40, the model management device 100 refers to the usage information of the training data stored in the usage information DB 30, and selects the training data specified by the data specifying information. Identify machine learning models trained on your data. Then, the model management device 100 updates the training data set by deleting the specified training
- the model management device 100 may be realized by a computing device such as a server, a personal computer (PC), a smartphone, or a tablet, and may have a hardware configuration as shown in FIG. 2, for example. That is, the model management device 100 includes a drive device 101, a storage device 102, a memory device 103, a processor 104, a user interface (UI) device 105, and a communication device 106 that are interconnected via a bus B.
- a computing device such as a server, a personal computer (PC), a smartphone, or a tablet
- the model management device 100 includes a drive device 101, a storage device 102, a memory device 103, a processor 104, a user interface (UI) device 105, and a communication device 106 that are interconnected via a bus B.
- UI user interface
- Programs or instructions for realizing various functions and processes to be described later in the model management device 100 may be stored in a removable storage medium such as a CD-ROM (Compact Disk-Read Only Memory) or a flash memory.
- a program or instruction is installed from the storage medium into the storage device 102 or the memory device 103 via the drive device 101.
- the program or instructions do not necessarily need to be installed from a storage medium, and may be downloaded from any external device via a network or the like.
- the storage device 102 is realized by a hard disk drive or the like, and stores installed programs or instructions as well as files, data, etc. used to execute the programs or instructions.
- the memory device 103 is realized by random access memory, static memory, etc., and when a program or instruction is started, reads the program or instruction, data, etc. from the storage device 102 and stores it.
- the storage device 102, the memory device 103, and the removable storage medium may be collectively referred to as a non-transitory tangible storage medium.
- the processor 104 may be realized by one or more CPUs (Central Processing Units), GPUs (Graphics Processing Units), processing circuits, etc. that may be configured from one or more processor cores, and may include memory.
- CPUs Central Processing Units
- GPUs Graphics Processing Units
- processing circuits etc. that may be configured from one or more processor cores, and may include memory.
- device 103 Various functions and processes of the model management device 100, which will be described later, are executed in accordance with stored programs, instructions, and data such as parameters necessary to execute the programs or instructions.
- the user interface (UI) device 105 may include input devices such as a keyboard, mouse, camera, and microphone, output devices such as a display, speaker, headset, and printer, and input/output devices such as a touch panel. It realizes an interface with the management device 100. For example, a user operates a GUI (Graphical User Interface) displayed on a display or a touch panel using a keyboard, a mouse, etc., to operate the model management device 100.
- GUI Graphic User Interface
- the communication device 106 is realized by various communication circuits that perform wired and/or wireless communication processing with communication networks such as external devices, the Internet, LAN (Local Area Network), and cellular networks.
- communication networks such as external devices, the Internet, LAN (Local Area Network), and cellular networks.
- model management device 100 may be realized by any other suitable hardware configuration.
- the model management device 100 refers to the training data usage information, identifies a machine learning model trained by the training data specified by the data identification information, and trains the identified machine learning model.
- the identified training data may be deleted from the used training data set, and processing (for example, retraining, etc.) for the machine learning model may be performed using the updated training data set.
- FIG. 3 is a block diagram showing the functional configuration of the model management device 100 according to an embodiment of the present disclosure.
- the model management device 100 includes a data specifying section 110, a model specifying section 120, and a processing section 130.
- the functional units of the data identification unit 110, the model identification unit 120, and the processing unit 130 may be configured such that one or more of the processors 104 uses a non-transitory tangible storage medium such as a storage device 102 and/or a memory device 103. This may be accomplished by executing one or more programs or instructions stored in the.
- the data specifying unit 110 acquires data specifying information. Specifically, upon acquiring identification information specifying training data from the terminal 40 or the like, the data specifying unit 110 identifies the corresponding identification information from the training data set stored in the training data DB 20 based on the acquired identification information. Identify the training data to use.
- the data identification information may indicate training data that was used to train a machine learning model managed by the model management device 100 but was found to be inappropriate after training.
- the data specific information may indicate training data that may degrade the performance of the machine learning model by using the training data.
- training data may be training data acquired through inappropriate acquisition methods, fabricated training data, training data acquired before the coronavirus pandemic, training data annotated by rogue contractors, etc. Good too.
- the data specifying information may indicate training data that may cause problems from a legal and/or ethical point of view if the training data is used.
- training data may include training data that is not authorized or licensed for commercial use, training data that was previously authorized or licensed but whose authorization or license has since become revoked, or training data that has not been authorized or licensed for commercial use.
- the data identification information may be identification information that identifies individual training data, but is not limited to this.
- the data specifying information is provided in the form of data specifying conditions (for example, indicating a specific acquisition location, acquisition time, annotation vendor, etc.) for specifying or extracting a plurality of training data stored in the training data DB 20. may be done.
- the data specifying unit 110 may specify a training data set corresponding to the acquired data specifying condition in the training data DB 20.
- the training data DB 20 holds detailed information for each training data as shown in FIG. It may be extracted as training data that corresponds to data specific conditions. For example, when the data specifying condition is the acquisition location "facility A", the data specifying unit 110 may extract training data #001, #005, #008 corresponding to the acquisition location "facility A”. Further, when the data specifying condition is personal information protection "permission”, the data specifying unit 110 may extract training data #001, #004, #006, #008 corresponding to personal information protection "permission”. .
- the model specifying unit 120 refers to the training data usage information and specifies a machine learning model trained using the training data specified by the data specifying information. Specifically, when the training data corresponding to the data specifying information acquired from the terminal 40 etc. is specified, the model specifying unit 120 accesses the usage information stored in the usage information DB 30 and Identify machine learning models trained using .
- the usage information may have a table-type data structure as shown in FIG. 4, and may have two columns: "model index” and "data index.”
- Model index identifies machine learning models #X1, #X2, . . . managed by the model management device 100.
- the "data index” identifies training data #001, #002, . . . stored in the training data DB 20, and indicates the training data used to train each machine learning model.
- the model specifying unit 120 may refer to the usage information and specify the model index #X1 corresponding to the data index #001. Further, when the training data #002 is specified by the data specifying information, the model specifying unit 120 may refer to the usage information to specify the model index #X2 corresponding to the data index #002.
- the processing unit 130 creates a new training data set by deleting the training data specified by the data identification information from the training data set used to train the machine learning model, and performs processing on the machine learning model. Execute. Specifically, when a machine learning model trained using the training data specified by the acquired data identification information is specified, the processing unit 130 determines the machine learning model used to train the specified machine learning model. The identified training data may be deleted from the training data set, and the training data set may be updated with the remaining training data. The processing unit 130 may then perform processing on the machine learning model based on the updated training dataset, for example, retraining the machine learning model using the updated training dataset. Good too.
- the processing unit 130 selects the training data sets #001, #004, #004, and The training data #001 may be deleted from the training data sets #006, #008, . . . and the machine learning model #1 may be retrained using the updated training data sets #004, #006, #008, .
- the processing unit 130 selects training data sets #002, #003, #005, #007 used to train machine learning model #X2. , #009,... may be deleted, and machine learning model #2 may be retrained using the updated training data sets #003, #005, #007, #009,... .
- the processing unit 130 may delete the training data specified by the data identification information from the training data DB 20, or may make it unusable. Thereby, it is possible to avoid using the training data that has become unusable, and it is also possible to release the storage area reserved for the training data that has become unusable.
- the processing unit 130 may determine whether the machine learning model is usable at a given point in time. Specifically, upon acquiring data specifying information indicating the expiration date of the training data, such as when the license M expires, the data specifying unit 110 refers to the detailed information in the training data DB 20 and selects “license M” as the license information. Specify training data #001, #004, #006, and #008 having the following. Since the identified training data #001, #004, #006, #008 will become unusable after the expiration date, the processing unit 130 will update the training data #001, #004, #006, #008 after the expiration date of the license M.
- the processing unit 130 retrains the machine learning model using the training data set from which training data #001, #004, #006, and #008 have been deleted, and after the expiration date of license M, the processing unit 130 The model may also be operated. This makes it possible to appropriately manage the operation of the machine learning model in accordance with the license deadline.
- a machine learning model trained using training data identified as inappropriate data can be retrained using a training dataset from which inappropriate data has been removed, and the machine It is possible to re-acquire a machine learning model that is suitable not only from the perspective of learning model performance but also from a legal and/or ethical perspective.
- model management processing is executed by the model management device 100 described above, and more specifically, one or more processors 104 of the model management device 100 execute one or more programs or programs stored in one or more memory devices 103. This may be accomplished by executing instructions.
- FIG. 6 is a flowchart showing model management processing according to an embodiment of the present disclosure.
- the model management device 100 acquires data specifying information.
- the model management device 100 may acquire data specifying information that specifies one or more inappropriate training data from the user via the terminal 40 or the like.
- the data specifying information may be identification information that identifies training data, or may be data specifying conditions that specify one or more training data.
- the model management device 100 refers to the detailed information in the training data DB 20 and determines that the "acquisition location" corresponds to "facility A.” Training data #001, #005, and #008 may be specified.
- the model management device 100 refers to the usage information and identifies a machine learning model trained using the identified training data. Specifically, the model management device 100 refers to the usage information stored in the usage information DB 30 and identifies the machine learning model trained using each training data identified in step S101. good. For example, the model management device 100 may specify two machine learning models #X1 and #X2 as machine learning models trained using training data #001, #005, and #008.
- step S104 the model management device 100 executes processing on the machine learning model.
- a machine learning model trained using training data identified as inappropriate data can be retrained using a training dataset from which inappropriate data has been removed, and the machine It is possible to re-acquire a machine learning model that is suitable not only from the perspective of learning model performance but also from a legal and/or ethical perspective.
- the model management system 10 further includes a data conversion device 50 and a conversion database (DB) 60.
- DB conversion database
- the data conversion device 50 executes data conversion to delete or anonymize personal information from the training data including personal information stored in the training data DB 20, and stores the converted training data in the conversion DB 60.
- the conversion DB 60 stores training data that has been converted by the data conversion device 50 to conceal or delete personal information from the data.
- the model management device 100 may use the converted training data stored in the conversion DB 60 instead of the training data stored in the training data DB 20 to train a machine learning model. Thereby, the model management device 100 can train a machine learning model using training data that does not include personal information.
- the conversion DB 60 may store detailed information of each training data stored in the training data DB 20 in association with the converted training data.
- the data specifying unit 110 when data specifying information for specifying inappropriate data is acquired, the data specifying unit 110 specifies the inappropriate training data identified by the data specifying information, and the model specifying unit 120 refers to the usage information to identify machine learning models that were trained using the identified inappropriate training data.
- the processing unit 130 identifies, in the conversion DB 60, the converted training dataset used to train the identified machine learning model, and deletes inappropriate training data from the identified converted training dataset. do. Furthermore, the processing unit 130 may retrain the machine learning model using the converted training data set from which inappropriate training data has been deleted.
- data conversion is performed to delete or anonymize personal information from training data that includes personal information, and the converted training data is used to train a machine learning model. Even if inappropriate training data is found while operating a machine learning model, delete the inappropriate training data and retrain the machine learning model using updated training data that does not include personal information. I can do it.
- a data identification unit that acquires data identification information; a model identifying unit that refers to training data usage information and identifies a first machine learning model trained using the first training data identified by the data identifying information; creating a second training dataset by removing the first training data from a first training dataset used to train the first machine learning model; a processing unit that executes processing for;
- a model management device having: (Additional note 2) The model management device according to supplementary note 1, wherein the usage information indicates an association between a machine learning model and training data used to train the machine learning model.
- the model management device includes: a data identification unit that acquires data identification information; a model identifying unit that refers to the usage information and identifies a first machine learning model trained using the first training data identified by the data identifying information; creating a second training dataset by removing the first training data from a first training dataset used to train the first machine learning model; a processing unit that executes processing for;
- a model management system with (Appendix 8) Obtaining data specific information; Referring to training data usage information, identifying a first machine learning model trained using the first training data identified by the data identification information; creating a second training dataset by removing the first training data from a first training dataset used to train the first machine learning model; performing processing on;
- a computer-implemented model management method comprising:
- Model management system 10
- Training data DB 30
- Usage information DB 40
- Terminal 50
- Data conversion device 60
- Conversion DB 100
- Model conversion device 110
- Data identification unit 120
- Model identification unit 130 Processing unit
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Bioethics (AREA)
- General Health & Medical Sciences (AREA)
- Computer Hardware Design (AREA)
- Computer Security & Cryptography (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Description
本開示を概略すると、このような不適切な訓練データが機械学習モデルを訓練するのに利用されたことが判明すると、本開示の一実施例によるモデル管理装置は、不適切な訓練データを利用して訓練された機械学習モデルを特定し、不適切な訓練データを訓練データセットから削除し、不適切な訓練データを含まない訓練データセットを利用して、特定した機械学習モデルを再訓練してもよい。
次に、図3~5を参照して、本開示の一実施例によるモデル管理装置100を説明する。本実施例によるモデル管理装置100は、訓練データの使用情報を参照して、データ特定情報によって特定された訓練データによって訓練された機械学習モデルを特定し、特定した機械学習モデルを訓練するのに使用された訓練データセットから、特定された訓練データを削除し、更新された訓練データセットによって当該機械学習モデルに対する処理(例えば、再訓練など)を実行してもよい。
次に、図6を参照して、本開示の一実施例によるモデル管理処理を説明する。当該モデル管理処理は、上述したモデル管理装置100によって実行され、より詳細には、モデル管理装置100の1つ以上のプロセッサ104が1つ以上のメモリ装置103に格納された1つ以上のプログラム又は指示を実行することによって実現されてもよい。図6は、本開示の一実施例によるモデル管理処理を示すフローチャートである。
次に、図7を参照して、本開示の他の実施例によるモデル管理システム10を説明する。本実施例によるモデル管理システム10は更に、データ変換装置50及び変換データベース(DB)60を有する。
(付記1)
データ特定情報を取得するデータ特定部と、
訓練データの使用情報を参照して、前記データ特定情報によって特定された第1の訓練データを使用して訓練された第1の機械学習モデルを特定するモデル特定部と、
前記第1の機械学習モデルを訓練するのに使用された第1の訓練データセットから前記第1の訓練データを削除することによって第2の訓練データセットを作成し、前記第1の機械学習モデルに対する処理を実行する処理部と、
を有する、モデル管理装置。
(付記2)
前記使用情報は、機械学習モデルと、前記機械学習モデルを訓練するのに使用された訓練データとの間の関連付けを示す、付記1に記載のモデル管理装置。
(付記3)
前記訓練データは、取得場所、取得時期、アノテーション情報、ライセンス情報、個人情報保護及び倫理情報の1つ以上と関連付けされる、付記1又は2に記載のモデル管理装置。
(付記4)
前記処理部は、前記第2の訓練データセットを使用して前記第1の機械学習モデルを再訓練する、付記1から3の何れか一項に記載のモデル管理装置。
(付記5)
前記訓練データは、データから個人情報を秘匿化又は削除することによって生成される、付記1から4の何れか一項に記載のモデル管理装置。
(付記6)
前記処理部は、所与の時点において前記第1の機械学習モデルが使用可能であるか判定する、付記1から5の何れか一項に記載のモデル管理装置。
(付記7)
訓練データを格納する訓練データデータベース(DB)と、
前記訓練データの使用情報を格納する使用情報データベース(DB)と、
前記訓練データDBと前記使用情報DBと通信接続されるモデル管理装置と、
を有し、
前記モデル管理装置は、
データ特定情報を取得するデータ特定部と、
前記使用情報を参照して、前記データ特定情報によって特定された第1の訓練データを使用して訓練された第1の機械学習モデルを特定するモデル特定部と、
前記第1の機械学習モデルを訓練するのに使用された第1の訓練データセットから前記第1の訓練データを削除することによって第2の訓練データセットを作成し、前記第1の機械学習モデルに対する処理を実行する処理部と、
を有する、モデル管理システム。
(付記8)
データ特定情報を取得することと、
訓練データの使用情報を参照して、前記データ特定情報によって特定された第1の訓練データを使用して訓練された第1の機械学習モデルを特定することと、
前記第1の機械学習モデルを訓練するのに使用された第1の訓練データセットから前記第1の訓練データを削除することによって第2の訓練データセットを作成し、前記第1の機械学習モデルに対する処理を実行することと、
を有する、コンピュータが実行するモデル管理方法。
20 訓練データDB
30 使用情報DB
40 端末
50 データ変換装置
60 変換DB
100 モデル変換装置
110 データ特定部
120 モデル特定部
130 処理部
Claims (8)
- データ特定情報を取得するデータ特定部と、
訓練データの使用情報を参照して、前記データ特定情報によって特定された第1の訓練データを使用して訓練された第1の機械学習モデルを特定するモデル特定部と、
前記第1の機械学習モデルを訓練するのに使用された第1の訓練データセットから前記第1の訓練データを削除することによって第2の訓練データセットを作成し、前記第1の機械学習モデルに対する処理を実行する処理部と、
を有する、モデル管理装置。 - 前記使用情報は、機械学習モデルと、前記機械学習モデルを訓練するのに使用された訓練データとの間の関連付けを示す、請求項1に記載のモデル管理装置。
- 前記訓練データは、取得場所、取得時期、アノテーション情報、ライセンス情報、個人情報保護及び倫理情報の1つ以上と関連付けされる、請求項1に記載のモデル管理装置。
- 前記処理部は、前記第2の訓練データセットを使用して前記第1の機械学習モデルを再訓練する、請求項1に記載のモデル管理装置。
- 前記訓練データは、データから個人情報を秘匿化又は削除することによって生成される、請求項1に記載のモデル管理装置。
- 前記処理部は、所与の時点において前記第1の機械学習モデルが使用可能であるか判定する、請求項1に記載のモデル管理装置。
- 訓練データを格納する訓練データデータベース(DB)と、
前記訓練データの使用情報を格納する使用情報データベース(DB)と、
前記訓練データDBと前記使用情報DBと通信接続されるモデル管理装置と、
を有し、
前記モデル管理装置は、
データ特定情報を取得するデータ特定部と、
前記使用情報を参照して、前記データ特定情報によって特定された第1の訓練データを使用して訓練された第1の機械学習モデルを特定するモデル特定部と、
前記第1の機械学習モデルを訓練するのに使用された第1の訓練データセットから前記第1の訓練データを削除することによって第2の訓練データセットを作成し、前記第1の機械学習モデルに対する処理を実行する処理部と、
を有する、モデル管理システム。 - データ特定情報を取得することと、
訓練データの使用情報を参照して、前記データ特定情報によって特定された第1の訓練データを使用して訓練された第1の機械学習モデルを特定することと、
前記第1の機械学習モデルを訓練するのに使用された第1の訓練データセットから前記第1の訓練データを削除することによって第2の訓練データセットを作成し、前記第1の機械学習モデルに対する処理を実行することと、
を有する、コンピュータが実行するモデル管理方法。
Priority Applications (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US18/871,512 US20250363412A1 (en) | 2022-06-09 | 2023-04-26 | Model management device, model management system, and model management method |
| JP2024526289A JP7648009B2 (ja) | 2022-06-09 | 2023-04-26 | モデル管理装置、モデル管理システム及びモデル管理方法 |
| EP23819531.7A EP4538939A4 (en) | 2022-06-09 | 2023-04-26 | MODEL MANAGEMENT DEVICE, MODEL MANAGEMENT SYSTEM, AND MODEL MANAGEMENT METHOD |
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2022-093639 | 2022-06-09 | ||
| JP2022093639 | 2022-06-09 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2023238544A1 true WO2023238544A1 (ja) | 2023-12-14 |
Family
ID=89118079
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/JP2023/016453 Ceased WO2023238544A1 (ja) | 2022-06-09 | 2023-04-26 | モデル管理装置、モデル管理システム及びモデル管理方法 |
Country Status (4)
| Country | Link |
|---|---|
| US (1) | US20250363412A1 (ja) |
| EP (1) | EP4538939A4 (ja) |
| JP (1) | JP7648009B2 (ja) |
| WO (1) | WO2023238544A1 (ja) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP7473274B1 (ja) | 2023-12-19 | 2024-04-23 | 株式会社フィードフォース | 情報処理システム、プログラム及び情報処理方法 |
| JP2025139757A (ja) * | 2024-03-13 | 2025-09-29 | キヤノン株式会社 | 情報処理装置およびその制御方法 |
| WO2025216298A1 (ja) * | 2024-04-12 | 2025-10-16 | 株式会社ちとせ研究所 | 解析用拡張データセット作成方法及び解析用拡張データセット作成用コンピュータプログラム |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2020102001A (ja) | 2018-12-21 | 2020-07-02 | ファナック株式会社 | 学習用データ確認支援装置、機械学習装置、故障予知装置 |
| WO2020202731A1 (ja) * | 2019-04-04 | 2020-10-08 | ソニー株式会社 | 情報処理システム及び情報処理方法 |
| JP2021157313A (ja) | 2020-03-25 | 2021-10-07 | 株式会社日立製作所 | 機械学習モデル運用管理システム、運用管理方法及びコンピュータプログラム |
| JP2022064214A (ja) * | 2020-10-13 | 2022-04-25 | 株式会社ブルーブックス | データ管理システム、データ管理方法、および機械学習データ管理プログラム |
| JP2022093639A (ja) | 2007-11-06 | 2022-06-23 | ピーティーシー セラピューティクス, インコーポレイテッド | ミトコンドリア病を治療するための4-(p-キノリル)-2-ヒドロキシブタンアミド誘導体 |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019187594A1 (ja) * | 2018-03-29 | 2019-10-03 | 日本電気株式会社 | 学習装置、学習方法および学習プログラム |
| EP3848836A1 (en) | 2020-01-07 | 2021-07-14 | Robert Bosch GmbH | Processing a model trained based on a loss function |
-
2023
- 2023-04-26 WO PCT/JP2023/016453 patent/WO2023238544A1/ja not_active Ceased
- 2023-04-26 US US18/871,512 patent/US20250363412A1/en active Pending
- 2023-04-26 JP JP2024526289A patent/JP7648009B2/ja active Active
- 2023-04-26 EP EP23819531.7A patent/EP4538939A4/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2022093639A (ja) | 2007-11-06 | 2022-06-23 | ピーティーシー セラピューティクス, インコーポレイテッド | ミトコンドリア病を治療するための4-(p-キノリル)-2-ヒドロキシブタンアミド誘導体 |
| JP2020102001A (ja) | 2018-12-21 | 2020-07-02 | ファナック株式会社 | 学習用データ確認支援装置、機械学習装置、故障予知装置 |
| WO2020202731A1 (ja) * | 2019-04-04 | 2020-10-08 | ソニー株式会社 | 情報処理システム及び情報処理方法 |
| JP2021157313A (ja) | 2020-03-25 | 2021-10-07 | 株式会社日立製作所 | 機械学習モデル運用管理システム、運用管理方法及びコンピュータプログラム |
| JP2022064214A (ja) * | 2020-10-13 | 2022-04-25 | 株式会社ブルーブックス | データ管理システム、データ管理方法、および機械学習データ管理プログラム |
Non-Patent Citations (1)
| Title |
|---|
| See also references of EP4538939A4 |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP7473274B1 (ja) | 2023-12-19 | 2024-04-23 | 株式会社フィードフォース | 情報処理システム、プログラム及び情報処理方法 |
| JP2025097875A (ja) * | 2023-12-19 | 2025-07-01 | 株式会社フィードフォース | 情報処理システム、プログラム及び情報処理方法 |
| JP2025139757A (ja) * | 2024-03-13 | 2025-09-29 | キヤノン株式会社 | 情報処理装置およびその制御方法 |
| WO2025216298A1 (ja) * | 2024-04-12 | 2025-10-16 | 株式会社ちとせ研究所 | 解析用拡張データセット作成方法及び解析用拡張データセット作成用コンピュータプログラム |
| JP2025161520A (ja) * | 2024-04-12 | 2025-10-24 | 株式会社ちとせ研究所 | 解析用拡張データセット作成方法及び解析用拡張データセット作成用コンピュータプログラム |
Also Published As
| Publication number | Publication date |
|---|---|
| EP4538939A4 (en) | 2025-11-05 |
| JP7648009B2 (ja) | 2025-03-18 |
| EP4538939A1 (en) | 2025-04-16 |
| JPWO2023238544A1 (ja) | 2023-12-14 |
| US20250363412A1 (en) | 2025-11-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116601644B (zh) | 使用分布式分类账提供可解释的机器学习模型结果 | |
| JP7648009B2 (ja) | モデル管理装置、モデル管理システム及びモデル管理方法 | |
| US11580951B2 (en) | Speaker identity and content de-identification | |
| US20220415203A1 (en) | Interface to natural language generator for generation of knowledge assessment items | |
| Park et al. | Understanding the representation and representativeness of age in AI data sets | |
| CN113961584A (zh) | 字段血缘分析方法、装置、电子设备及存储介质 | |
| CN103827899B (zh) | 数据中心资源分配 | |
| US10496744B2 (en) | Domain-specific lexically-driven pre-parser | |
| CN102163215A (zh) | 来自文档概述的问题概念的本体分类 | |
| CN107145395A (zh) | 用于处理任务的方法和装置 | |
| JP2013533993A (ja) | コンテキスト認識のデータを保護するための方法、システム、及びコンピュータ・プログラム | |
| JP2021500692A (ja) | 系図エンティティ解決システムおよび方法 | |
| KR102655929B1 (ko) | 생성형 ai 모델을 활용한 대화형 텍스트 데이터 생성 시스템 및 그 방법 | |
| JP2024524094A (ja) | データガバナンスシステム及び方法 | |
| JP2016031733A (ja) | 推論容易性算出プログラム、装置、及び方法 | |
| US20130006683A1 (en) | System and method of managing testing for a healthcare systems transition | |
| WO2022234734A1 (ja) | データサービス提供方法およびデータサービス提供システム | |
| CN114942749B (zh) | 审批系统开发方法、装置、设备及存储介质 | |
| CN112259182A (zh) | 一种电子病历的生成方法和装置 | |
| CN112416983B (zh) | 一种数据处理方法及装置、计算机可读存储介质 | |
| CN104424525B (zh) | 辅助确定项目范围的方法和装置 | |
| US20130054301A1 (en) | Process refinement | |
| CN115659182B (zh) | 一种模型更新方法、装置及设备 | |
| JP7760566B2 (ja) | 情報処理装置、情報処理方法、及び情報処理プログラム | |
| CN120447932B (zh) | 一种数据库版本升级管理方法、装置、设备以及存储介质 |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 23819531 Country of ref document: EP Kind code of ref document: A1 |
|
| ENP | Entry into the national phase |
Ref document number: 2024526289 Country of ref document: JP Kind code of ref document: A |
|
| WWE | Wipo information: entry into national phase |
Ref document number: 18871512 Country of ref document: US |
|
| WWE | Wipo information: entry into national phase |
Ref document number: 2023819531 Country of ref document: EP |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| ENP | Entry into the national phase |
Ref document number: 2023819531 Country of ref document: EP Effective date: 20250109 |
|
| WWP | Wipo information: published in national office |
Ref document number: 2023819531 Country of ref document: EP |
|
| WWP | Wipo information: published in national office |
Ref document number: 18871512 Country of ref document: US |