WO2024189831A1 - 学習装置、学習方法、および学習プログラム - Google Patents

学習装置、学習方法、および学習プログラム Download PDF

Info

Publication number
WO2024189831A1
WO2024189831A1 PCT/JP2023/010094 JP2023010094W WO2024189831A1 WO 2024189831 A1 WO2024189831 A1 WO 2024189831A1 JP 2023010094 W JP2023010094 W JP 2023010094W WO 2024189831 A1 WO2024189831 A1 WO 2024189831A1
Authority
WO
WIPO (PCT)
Prior art keywords
series data
time series
class
data
teacher
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
Application number
PCT/JP2023/010094
Other languages
English (en)
French (fr)
Japanese (ja)
Inventor
竜馬 大内
真 寺尾
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
NEC Corp
Original Assignee
NEC Corp
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 NEC Corp filed Critical NEC Corp
Priority to JP2025506361A priority Critical patent/JPWO2024189831A1/ja
Priority to PCT/JP2023/010094 priority patent/WO2024189831A1/ja
Publication of WO2024189831A1 publication Critical patent/WO2024189831A1/ja
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning

Definitions

  • the present invention relates to a learning device, a learning method, and a learning program.
  • the class yi may be one class or multiple classes.
  • the correct class label yi is generally given as training data for all time series data xi.
  • the class label in this case is called a full label.
  • the labeling cost of a full label is very high.
  • Non-Patent Document 1 describes a method for learning a video action segment recognition task, which is an example of a time-series data segment recognition task, using weak labels.
  • a model is first trained using timestamp-type labels.
  • pseudo labels are assigned to areas near the data at the time to which the label is assigned, and these are used together for learning.
  • a pseudo label is a label that is assigned pseudo-wise to data at a time to which no label is assigned.
  • One aspect of the present invention has been made in consideration of the above problems, and one example of the purpose of the present invention is to enable learning of a highly accurate machine learning model that infers into which class data at each time point in time series data is classified while reducing the labeling cost.
  • a learning device is a learning device that uses multiple teacher time series data to machine-learn a machine learning model that infers into which class data at each time point in time series data is classified, and includes a class matching unit that matches each teacher time series data with the class indicated by the label assigned to the data included in the teacher time series data, and a class matching unit that matches at least one teacher time series data with other teacher time series data based on the similarity between the teacher time series data.
  • the system includes a class propagation unit that associates at least some of the classes assigned to each of the teacher time series data, a pseudo label assignment unit that assigns a pseudo label indicating the class into which the machine learning model has classified data that is not assigned to the label included in the teacher time series data, and a learning unit that trains the machine learning model by machine learning using the multiple teacher time series data including data to which the pseudo label has been assigned, and the pseudo label assignment unit limits the pseudo label to be assigned to data included in the teacher time series data based on the class associated with each of the teacher time series data.
  • a learning method is a learning method for machine learning a machine learning model that infers into which class data at each time point of time series data is classified, using a plurality of teacher time series data, wherein some data included in the plurality of teacher time series data are assigned labels indicating the classes, and the method includes a class matching process for matching each teacher time series data with the class indicated by the label assigned to the data included in the teacher time series data, and a class matching process for matching at least one teacher time series data with other teacher time series data based on the similarity between the teacher time series data.
  • the method executes a class propagation process that associates at least some of the classes associated with each of the teacher time series data, a pseudo label assignment process that assigns pseudo labels indicating the classes into which the machine learning model has classified data that is not assigned to the labeled data included in the teacher time series data, and a learning process that trains the machine learning model by machine learning using the multiple teacher time series data including data to which the pseudo labels have been assigned, and the pseudo label assignment process limits the pseudo labels to be assigned to data included in the teacher time series data based on the classes associated with each of the teacher time series data.
  • a learning program is a learning program that causes a computer to machine-learn a machine learning model that infers into which class data at each time point of time series data is classified, using a plurality of teacher time series data, and some data included in the plurality of teacher time series data are given labels indicating the classes, and the program includes a class matching process that matches each teacher time series data with the class indicated by the label given to the data included in the teacher time series data, and a class matching process that matches at least one teacher time series data with other teacher time series data based on the similarity between the teacher time series data.
  • the system executes a class propagation process that assigns at least a portion of the classes associated with each of the teacher time series data to the data that is not assigned a label, a pseudo label assignment process that assigns a pseudo label indicating the class into which the machine learning model has classified the data to data included in the teacher time series data that is not assigned a label, and a learning process that trains the machine learning model by machine learning using the multiple teacher time series data that include data to which the pseudo label has been assigned, and the pseudo label assignment process limits the pseudo label to be assigned to data included in the teacher time series data based on the class associated with each of the teacher time series data.
  • the learning device 1 uses multiple teacher time series data to perform machine learning on a machine learning model that infers which class data at each time of the time series data is classified into.
  • a label indicating a class is assigned to some data included in the multiple teacher time series data.
  • the label indicating a class may be, for example, one label assigned to one piece of data at each time of the time series data, or multiple labels may be assigned.
  • the multiple teacher time series data may include, for example, multiple independent pieces of data, or may include multiple pieces of time series data that are related to each other and are generated by dividing one piece of time series data into multiple pieces.
  • the time series data is either fully labeled time series data, partially labeled time series data, or completely unlabeled time series data.
  • FIG. 1 is a block diagram showing the configuration of the learning device 1.
  • the learning device 1 includes a class matching unit 11, a class propagation unit 12, a pseudo label assignment unit 13, and a learning unit 14.
  • the class propagation unit 12 associates at least one teacher time series data with at least a portion of the classes associated with other teacher time series data based on the similarity between the teacher time series data.
  • the similarity indicates how similar the characteristics of each time series data are to each other.
  • the features of each time series data unit are represented by feature amounts.
  • the feature amount in the video is, for example, the average of the feature amounts of all frames. In the space representing the feature amounts, the closer the positions of the feature amounts are, the higher the similarity is determined to be.
  • the class propagation unit 12 assumes that time series data with sufficiently high similarity to each other have similar classes, and associates all or part of the classes associated with one time series data with the other time series data.
  • the class propagation unit 12 may select K classes (K is a natural number and is less than or equal to the total number of time series data) in descending order of similarity to time series data whose classes in the time series data are known, and assign all or some of the classes in the time series data to the time series data.
  • the label of that class may be considered reliable and may be considered a valid class only if the same class label is assigned from multiple labeled time series data.
  • the class propagation unit 12 when the class propagation unit 12 focuses on time series data whose class within the time series data is unknown, and there are multiple pieces of time series data whose class within the time series data is known and has a similarity, the known class in the time series data with the largest total number may be assigned to the time series data whose class within the time series data is unknown.
  • classes within the time series data may be weighted by the similarity for time series data whose similarity is sufficiently close.
  • the class propagation unit 12 may further assign the class within the propagated time series data to other time series data.
  • the pseudo-labeling unit 13 assigns pseudo-labels to unlabeled data included in each teacher time-series data, indicating the class into which the machine learning model has classified the data.
  • pseudo-labels based on data that has already been labeled can be assigned to both unlabeled data and data that has already been labeled.
  • the pseudo label assignment unit 13 restricts the pseudo labels to be assigned to data included in the teacher time series data based on the class associated with each teacher time series data.
  • the pseudo labels to be assigned are restricted based on the class already associated with the time series data.
  • the conditions for restriction include, for example, the constraint conditions in the exemplary embodiment 2 described below.
  • the learning unit 14 trains a machine learning model using multiple training time-series data, including data to which pseudo-labels have been assigned.
  • the learning device 1 configured as above executes a learning method S1 according to this exemplary embodiment.
  • Learning method S1 uses multiple teacher time series data to machine-learn a machine learning model that infers into which class data at each time point in the time series data is classified. Some data included in the multiple teacher time series data is given a label indicating the class.
  • FIG. 2 is a flow diagram showing the flow of the learning method S1.
  • the learning method S1 includes a class matching step S11, a class propagation step S12, a pseudo-label assignment step S13, and a learning step S14.
  • the class matching step S11 the class matching unit 11 matches each teacher time series data with a class indicated by a label assigned to data included in the teacher time series data.
  • the class propagation step S12 the class propagation unit 12 matches at least one teacher time series data with at least a part of a class associated with other teacher time series data based on the similarity between the teacher time series data.
  • the pseudo-label assignment unit 13 assigns a pseudo label indicating a class into which the machine learning model has classified the data to data that is not assigned a label included in the teacher time series data.
  • the pseudo label assignment step S13 limits the pseudo labels to be assigned to the data included in the teacher time series data based on the class associated with each teacher time series data.
  • the learning step S14 the learning unit 14 trains a machine learning model by using multiple teacher time series data including data to which pseudo labels have been assigned.
  • the learning device 1 and the learning method S1 according to this exemplary embodiment can prevent the assignment of pseudo labels of wrong classes, such as classes that do not exist in the teacher time-series data. As a result, the number and variety of assigned pseudo labels increases, which is expected to result in high inference accuracy.
  • Exemplary embodiment 2 A second exemplary embodiment of the present invention will be described in detail with reference to the drawings. Note that components having the same functions as those described in the first exemplary embodiment are denoted by the same reference numerals, and the description thereof will be omitted as appropriate.
  • FIG. 3 is a block diagram showing the functional configuration of the learning device 10.
  • the learning device 10 includes a control unit 110 and a storage unit 120.
  • the control unit 110 controls each unit of the learning device 10.
  • the control unit 110 includes a class matching unit 11, a class propagation unit 12, a pseudo label assignment unit 13, a learning unit 14, an inference unit 15, a feature amount calculation unit 16, a similarity calculation unit 17, and a constraint condition assignment unit 18.
  • the storage unit 120 stores various data used by the control unit 110. For example, the storage unit 120 stores teacher time series data TD and a machine learning model MM.
  • the class matching unit 11 matches each teacher time series data TD with a class indicated by a label assigned to data included in the teacher time series data TD.
  • Matching a class refers to matching a class to the entirety of each time series data.
  • the class matching unit 11 assigns to the time series data a class indicated by a label assigned to data included in the time series data. For example, when the time series data is a video, this refers to assigning to each video a class indicated by a label assigned to a frame within the video. In one aspect, when a class corresponding to time series data is directly specified, the class matching unit 11 may assign the class to the time series data.
  • the class propagation unit 12 associates at least one teacher time series data TD with at least a portion of the classes associated with other teacher time series data TD based on the similarity between the teacher time series data TD.
  • the similarity indicates how similar the characteristics of each time series data are to each other.
  • the features of each time series data unit are represented by feature amounts.
  • the feature amount in the video is, for example, the average of the feature amounts of all frames. In the space representing the feature amounts, the closer the positions of the feature amounts are, the higher the similarity is determined to be.
  • the class propagation unit 12 assumes that time series data with sufficiently high similarity to each other have similar classes, and associates all or part of the classes associated with one time series data with the other time series data.
  • the class propagation unit 12 may associate all classes associated with the first time series data with second time series data whose similarity to the first time series data is equal to or greater than a predetermined threshold.
  • the class propagation unit 12 may also associate some of the classes associated with the first time series data with second time series data whose similarity to the first time series data is equal to or greater than a predetermined threshold.
  • a feature may be generated for each class of time series data, and the class may be propagated between time series data whose similarity of the feature for the class is equal to or greater than a predetermined threshold.
  • class A when there is time series data having classes A, B, and C, if the similarity of the feature for class A is sufficiently high, class A may be assigned, and if the similarity of the feature for class C is sufficiently low, class C may not be assigned.
  • classes may be assigned only to a portion of time series data having multiple classes.
  • the feature for each class may be calculated, for example, by using a machine learning model that receives input data of the time series data and outputs a feature, and that is machine-learned so that the output feature is large when data to which each class is assigned is input.
  • the class propagation unit 12 may assign all or some of the classes in the time series data to the time series data whose classes in the time series data are known, by selecting K classes (K is a natural number and is less than or equal to the total number of time series data) in order of similarity.
  • the label of that class may be considered reliable and may be considered a valid class only if the same class label is assigned from multiple labeled time series data.
  • the class propagation unit 12 when the class propagation unit 12 focuses on time series data whose class within the time series data is unknown, and there are multiple pieces of time series data whose class within the time series data is known and has a similarity, the known class in the time series data with the largest total number may be assigned to the time series data whose class within the time series data is unknown.
  • classes within the time series data may be weighted by the similarity for time series data whose similarity is sufficiently close.
  • the class propagation unit 12 may further assign the class within the propagated time series data to other time series data.
  • the pseudo label assignment unit 13 assigns pseudo labels to unlabeled data included in the teacher time series data TD, indicating the class into which the machine learning model MM has classified the data.
  • pseudo labels based on data that has already been labeled can be assigned to both unlabeled data and data that has already been labeled.
  • the pseudo label assignment unit 13 restricts the pseudo labels to be assigned to the data included in the teacher time series data TD based on the class associated with each teacher time series data TD.
  • the pseudo labels to be assigned are restricted based on the class already associated with the time series data.
  • the conditions for restriction include, for example, the constraint conditions in the exemplary embodiment 2 described below.
  • the learning unit 14 trains the machine learning model MM by using a plurality of teacher time series data TD including data to which pseudo labels have been assigned.
  • the learning unit 14 may further include a configuration for calculating a loss using, for example, the labels originally assigned in the teacher time series data TD, the pseudo labels assigned to the teacher time series data TD, and the result of inference as inputs, and updating the parameters of the machine learning model MM using the loss as input.
  • the loss refers to the magnitude of the deviation between the labels originally assigned in the teacher time series data TD or the pseudo labels assigned to the teacher time series data TD, and the result of inference.
  • the inference unit 15 infers into which class the data at each time point in the teacher time series data TD is classified.
  • the feature calculation unit 16 calculates features for each piece of teacher time series data TD on a time series data basis.
  • the features may be the output result of a pre-trained model, color features, or meta information.
  • Meta information may be, for example, the time at which the time series data was acquired, or the location at which the time series data was acquired.
  • the feature may use the angle of view of the camera that acquired the video.
  • feature amounts may be calculated from values that represent the features of each piece of data at each time of the time series data, which are output from the intermediate and final layers of the neural network. Furthermore, feature amounts may be calculated after performing a pooling process such as averaging on the output values. Furthermore, pooling may be performed by weighting using a prediction score or the like. Furthermore, the output values may be passed through yet another neural network, and, for example, metric learning or contrastive learning may be performed in that space.
  • the feature calculation unit 16 may calculate the feature from the time ratio of the time series data section estimated from the inference result (for example, if the time series data section is an action section in a video, what is the time ratio of each action).
  • the feature may be calculated so that, for example, if the action time ratios in the videos are similar, it can be determined that the similarity between the videos is high.
  • the similarity calculation unit 17 uses the features to calculate the similarity between the teacher time series data TD.
  • the similarity calculation may use cosine similarity, Euclidean distance, Manhattan distance (L1 norm), or Kullback-Leibler divergence (K-L divergence).
  • the constraint condition assigning unit 18 assigns constraint conditions that limit the class of the pseudo label to the class of the label that originally exists in the teacher time series data TD, or the class of the label in the teacher time series data TD that is obtained by being assigned by the class propagation unit.
  • the constraint condition assigning unit 18 may set a constraint condition that restricts the assignment of pseudo labels to only classes in the teacher time-series data TD and does not assign pseudo labels to other classes.
  • the constraint condition assignment unit 18 may assign different pseudo label thresholds to classes that are the same as the classes of the labels in the teacher time series data TD and other classes.
  • the constraint condition assigning unit 18 may change the constraint conditions on the teacher time-series data TD depending on the progress of the machine learning. Examples of the changes include removing or relaxing the constraint conditions.
  • Flow of learning method S10 The learning device 10 configured as above executes a learning method S10 according to this exemplary embodiment.
  • the flow of the learning method S10 will be described with reference to Fig. 4.
  • Fig. 4 is a flow diagram showing the flow of the learning method S10. As shown in Fig. 4, the learning method S10 includes steps S101 to S108.
  • the class matching unit 11 matches each teacher time series data TD with a class indicated by a label assigned to the data included in the teacher time series data TD.
  • inference step S102 the inference unit 15 infers into which class the data at each time point in the teacher time series data TD is classified.
  • the feature calculation unit 16 calculates the feature of each time series data unit for each teacher time series data TD.
  • the similarity calculation unit 17 uses the features to calculate the similarity between the teacher time series data TD.
  • the class propagation unit 12 associates at least one teacher time series data TD with at least a portion of the classes associated with other teacher time series data TD based on the similarity between the teacher time series data TD.
  • the constraint condition assignment unit 18 assigns constraint conditions that limit the class of the pseudo label to the class of the label that originally exists in the teacher time series data TD, or the class of the label in the teacher time series data TD obtained by being assigned by the class propagation unit.
  • the pseudo label assignment unit 13 assigns a pseudo label indicating the class into which the machine learning model MM has classified the data to the unlabeled data included in the teacher time series data TD for each teacher time series data TD. Note that the pseudo label assignment step S107 limits the pseudo label to be assigned to the data included in the teacher time series data TD based on the class associated with each teacher time series data TD.
  • the learning unit 14 uses multiple pieces of teacher time series data TD, including data to which pseudo labels have been assigned, to train the machine learning model MM.
  • the learning unit S108 may further include a configuration for calculating a loss using, for example, the labels originally assigned in the teacher time series data TD, the pseudo labels assigned to the teacher time series data TD, and the result of the inference as inputs, and updating the parameters of the machine learning model MM using the loss as input.
  • the learning device 10 and the learning method S10 according to this exemplary embodiment impose constraints on the pseudo labels, thereby making it possible to prevent the assignment of pseudo labels of wrong classes, such as classes that do not exist in the teacher time-series data TD. As a result, the number and variety of assigned pseudo labels increases, which is expected to result in high inference accuracy.
  • the feature acquisition unit 21 acquires the features of the teacher time series data TD.
  • the clustering unit 22 clusters the features obtained by the feature acquisition unit 21.
  • k-means or TW-FINCH may be used as the clustering method.
  • the data selection unit 23 selects data from near the center of each cluster using the clustering results obtained by the clustering unit 22.
  • the data selection unit 23 selects data from near the center of the cluster that represents each feature in each cluster divided according to the features of the teacher time-series data TD, and obtains the time of the selected data.
  • the label acquisition unit 24 acquires a label to be assigned to the data at each time in the teacher time series data TD, which corresponds to the time obtained by the data selection unit 23.
  • the label acquired by the label acquisition unit 24 may be assigned manually by a person to the data at each time in each teacher time series data TD.
  • the class matching unit 25, the class propagation unit 26, the pseudo label assignment unit 27, and the learning unit 28 have the same functions as the class matching unit 11, the class propagation unit 12, the pseudo label assignment unit 13, and the learning unit 14 described in the exemplary embodiment 1, and therefore will not be described here.
  • Flow of learning method S20 The learning device 20 configured as above executes a learning method S20 according to this exemplary embodiment.
  • the flow of the learning method S20 will be described with reference to Fig. 6.
  • Fig. 6 is a flow chart showing the flow of the learning method S20. As shown in Fig. 6, the learning method S20 includes steps S201 to S208.
  • feature acquisition step S201 the feature acquisition unit 21 acquires the features of the teacher time-series data TD.
  • the clustering unit 22 clusters the features obtained in the feature acquisition step S201.
  • the data selection unit 23 selects data from near the center of each cluster using the clustering results obtained in the clustering step S202.
  • the label acquisition unit 24 acquires a label to be assigned to the data at each time in the teacher time series data TD, which corresponds to the time obtained in the data selection step S203.
  • the label acquired in the label acquisition step S204 is assigned to the data at each time in each teacher time series data TD, for example, manually, before proceeding to the processing of the class matching step S205 and subsequent steps.
  • class matching step S205, class propagation step S206, pseudo label assignment step S207, and learning step S208 have the same processing as the class matching step S11, class propagation step S12, pseudo label assignment step S13, and learning step S14 described in exemplary embodiment 1, so their explanations are omitted.
  • the learning device 20 and learning method S20 make it possible to acquire data at each time of the teacher time series data TD that have different characteristics from each other in the teacher time series data TD.
  • the learning device 20 and learning method S20 make it possible to acquire data at each time of the teacher time series data TD that have different characteristics from each other in the teacher time series data TD.
  • the data at each time of the teacher time series data TD acquired in this way as a target for labeling, it becomes possible to reduce the cost of searching for data to be labeled within the teacher time series data TD.
  • each device may be realized by hardware such as an integrated circuit (IC chip), or may be realized by software.
  • each device is realized, for example, by a computer that executes instructions of a program, which is software that realizes each function.
  • a computer that executes instructions of a program, which is software that realizes each function.
  • An example of such a computer (hereinafter referred to as computer C) is shown in Figure 7.
  • Computer C has at least one processor C1 and at least one memory C2.
  • Memory C2 stores program P for operating computer C as each device.
  • processor C1 reads and executes program P from memory C2, thereby realizing each function of each device.
  • the processor C1 may be, for example, a CPU (Central Processing Unit), GPU (Graphic Processing Unit), DSP (Digital Signal Processor), MPU (Micro Processing Unit), FPU (Floating point number Processing Unit), PPU (Physics Processing Unit), TPU (Tensor Processing Unit), quantum processor, microcontroller, or a combination of these.
  • the memory C2 may be, for example, a flash memory, HDD (Hard Disk Drive), SSD (Solid State Drive), or a combination of these.
  • Computer C may further include a RAM (Random Access Memory) for expanding program P during execution and for temporarily storing various data.
  • Computer C may further include a communications interface for sending and receiving data to and from other devices.
  • Computer C may further include an input/output interface for connecting input/output devices such as a keyboard, mouse, display, and printer.
  • the program P can also be recorded on a non-transitory, tangible recording medium M that can be read by the computer C.
  • a recording medium M can be, for example, a tape, a disk, a card, a semiconductor memory, or a programmable logic circuit.
  • the computer C can obtain the program P via such a recording medium M.
  • the program P can also be transmitted via a transmission medium.
  • a transmission medium can be, for example, a communications network or broadcast waves.
  • the computer C can also obtain the program P via such a transmission medium.
  • a learning device that performs machine learning to generate a machine learning model that infers into which class data at each time point of the time series data is classified, using a plurality of teacher time series data, comprising: A label indicating the class is assigned to some data included in the plurality of teacher time-series data; a class association unit that associates, with each teacher time series data, a class indicated by the label assigned to data included in the teacher time series data; a class propagation unit that associates at least one of the teacher time series data with at least a part of the classes associated with other teacher time series data based on the similarity between the teacher time series data; a pseudo-labeling unit that assigns a pseudo-label indicating a class into which the machine learning model has classified data to data that is not assigned the label and is included in the teacher time-series data for each teacher time-series data; a learning unit that uses the plurality of teacher time-series data including the data to which the pseudo-label is assigned to learn the machine learning model;
  • (Appendix 2) an inference unit that infers into which class data at each time point of the teacher time series data is classified;
  • a feature amount calculation unit that calculates a feature amount of each of the teacher time series data units;
  • a similarity calculation unit that calculates the similarity between the teacher time-series data by using the feature amount;
  • a constraint condition assigning unit that assigns a constraint condition to restrict the class of the pseudo label to the class of the label originally present in the teacher time-series data or the class of the label in the teacher time-series data obtained by assigning the class by the class propagation unit;
  • the learning unit is Calculating a loss using the label originally assigned to the teacher time-series data, the pseudo-label assigned to the teacher time-series data, and the result of the inference as input; updating parameters of the machine learning model using the loss as an input; 2.
  • a learning device as described in claim 1.
  • a feature acquisition unit for acquiring features of the teacher time series data for acquiring features of the teacher time series data; a clustering unit that clusters the feature amounts obtained by the feature amount acquisition unit; a data selection unit that selects data from near the center of each cluster using the clustering results obtained by the clustering unit; a label acquisition unit that acquires the label to be assigned to data at each time of the teacher time-series data corresponding to the time obtained by the data selection unit, 3.
  • a learning device according to claim 1 or 2.
  • the feature is an output result of a pre-trained model, a color feature, or meta information. 4.
  • a learning device according to claim 2 or 3.
  • the meta information is the acquisition time of the time series data or the acquisition location of the time series data. 5.
  • a learning method for machine learning a machine learning model that infers into which class data at each time point of time series data is classified, using a plurality of teacher time series data comprising: A label indicating the class is assigned to some data included in the plurality of teacher time-series data; A class matching process for matching each teacher time series data with a class indicated by the label assigned to the data included in the teacher time series data; A class propagation process for associating at least a part of classes associated with other teacher time series data with at least one teacher time series data based on the similarity between the teacher time series data; a pseudo-labeling process for assigning a pseudo-label indicating a class into which the machine learning model has classified data to data that is not assigned a label included in each teacher time series data; A learning process of machine learning the machine learning model using the plurality of training time-series data including the data to which the pseudo-labels are assigned; The learning method, in which the pseudo label assignment process limits the pseudo labels to be assigned to data included in each
  • a learning program for performing machine learning on a machine learning model that infers into which class data at each time point of time series data is classified, using a plurality of teacher time series data comprising: A label indicating the class is assigned to some data included in the plurality of teacher time-series data; A class matching process for matching each teacher time series data with a class indicated by the label assigned to the data included in the teacher time series data; A class propagation process for associating at least a part of classes associated with other teacher time series data with at least one teacher time series data based on the similarity between the teacher time series data; a pseudo-labeling process for assigning a pseudo-label indicating a class into which the machine learning model has classified data to data that is not assigned a label included in each teacher time series data; a learning process for learning the machine learning model by using the plurality of training time-series data including the data to which the pseudo-labels are assigned; A learning program in which the pseudo label assignment process limits the pseudo labels
  • a learning device that performs machine learning to generate a machine learning model that infers into which class data at each time point of the time series data is classified, using a plurality of teacher time series data, comprising: A label indicating the class is assigned to some data included in the plurality of teacher time-series data; A class matching process for matching each teacher time series data with a class indicated by the label assigned to the data included in the teacher time series data; A class propagation process for associating at least a part of classes associated with other teacher time series data with at least one teacher time series data based on the similarity between the teacher time series data; a pseudo-labeling process for assigning a pseudo-label indicating a class into which the machine learning model has classified data to data that is not assigned a label included in each teacher time series data; A learning process of machine learning the machine learning model using the plurality of training time-series data including the data to which the pseudo-labels are assigned; The pseudo label assignment process limits the pseudo labels to
  • the learning device may further include a memory, and the memory may store a program for causing the processor to execute the class matching process, the class propagation process, the pseudo-labeling process, and the learning process.
  • the program may also be recorded on a computer-readable, non-transitory, tangible recording medium.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Medical Informatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Artificial Intelligence (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Image Analysis (AREA)
PCT/JP2023/010094 2023-03-15 2023-03-15 学習装置、学習方法、および学習プログラム Ceased WO2024189831A1 (ja)

Priority Applications (2)

Application Number Priority Date Filing Date Title
JP2025506361A JPWO2024189831A1 (https=) 2023-03-15 2023-03-15
PCT/JP2023/010094 WO2024189831A1 (ja) 2023-03-15 2023-03-15 学習装置、学習方法、および学習プログラム

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/JP2023/010094 WO2024189831A1 (ja) 2023-03-15 2023-03-15 学習装置、学習方法、および学習プログラム

Publications (1)

Publication Number Publication Date
WO2024189831A1 true WO2024189831A1 (ja) 2024-09-19

Family

ID=92754720

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2023/010094 Ceased WO2024189831A1 (ja) 2023-03-15 2023-03-15 学習装置、学習方法、および学習プログラム

Country Status (2)

Country Link
JP (1) JPWO2024189831A1 (https=)
WO (1) WO2024189831A1 (https=)

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2019159576A (ja) * 2018-03-09 2019-09-19 富士通株式会社 学習プログラム、学習方法および学習装置
JP2021196921A (ja) * 2020-06-16 2021-12-27 株式会社日立製作所 モデル運用支援システム及び方法

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2019159576A (ja) * 2018-03-09 2019-09-19 富士通株式会社 学習プログラム、学習方法および学習装置
JP2021196921A (ja) * 2020-06-16 2021-12-27 株式会社日立製作所 モデル運用支援システム及び方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
SAKAGUCHI SHOKI, AMAGASAKI MOTOKI, KIYAMA MASATO, OKAMOTO TOSHIAKI: "F-014: A study on person identification using multiple surveillance cameras", THE 21TH FORUM ON INFORMATION TECHNOLOGY, IEICE, vol. 2, 30 August 2022 (2022-08-30), pages 373 - 376, XP093210355 *

Also Published As

Publication number Publication date
JPWO2024189831A1 (https=) 2024-09-19

Similar Documents

Publication Publication Date Title
US11574147B2 (en) Machine learning method, machine learning apparatus, and computer-readable recording medium
US10558885B2 (en) Determination method and recording medium
CN110362677B (zh) 文本数据类别的识别方法及装置、存储介质、计算机设备
US10402627B2 (en) Method and apparatus for determining identity identifier of face in face image, and terminal
US10373028B2 (en) Pattern recognition device, pattern recognition method, and computer program product
CN113657260A (zh) 一种弱监督时序行为定位方法
US10614312B2 (en) Method and apparatus for determining signature actor and identifying video based on probability of appearance of signature actor
CN102473409B (zh) 声音空间的基准模型适应装置、集成电路以及av设备
US20150363667A1 (en) Recognition device and method, and computer program product
US20140257810A1 (en) Pattern classifier device, pattern classifying method, computer program product, learning device, and learning method
KR101514551B1 (ko) 환경 변화에 강인한 멀티모달 사용자 인식
JP7389389B2 (ja) 処理装置、処理方法および処理プログラム
JP2014228995A (ja) 画像特徴学習装置、画像特徴学習方法及びプログラム
US20200019875A1 (en) Parameter calculation device, parameter calculation method, and non-transitory recording medium
US10915794B2 (en) Neural network classification through decomposition
JP7559943B2 (ja) 情報処理装置、情報処理方法及びプログラム
WO2024189831A1 (ja) 学習装置、学習方法、および学習プログラム
US11869492B2 (en) Anomaly detection system and method using noise signal and adversarial neural network
JP6947460B1 (ja) プログラム、情報処理装置、及び方法
JP7680660B2 (ja) 情報処理装置、情報処理方法、およびプログラム
US20240221368A1 (en) Information processing apparatus, information processing method, and recording medium
US20240303543A1 (en) Model training method and model training apparatus
EP4675576A1 (en) System and method for generating a federated learning based model for improved video analytics
WO2024189833A1 (ja) 情報処理装置、情報処理方法、及び情報処理プログラム
WO2024189832A1 (ja) 情報処理装置、情報処理方法、及び情報処理プログラム

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: 23927451

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2025506361

Country of ref document: JP

Kind code of ref document: A

WWE Wipo information: entry into national phase

Ref document number: 2025506361

Country of ref document: JP

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 23927451

Country of ref document: EP

Kind code of ref document: A1