Search

US-12619869-B2 - Learning apparatus, learning method, and a non-transitory computer-readable storage medium

US12619869B2US 12619869 B2US12619869 B2US 12619869B2US-12619869-B2

Abstract

A learning apparatus according to the present application includes: a dividing unit that divides predetermined learning data features of which are to be learned by a model by training, into a plurality of sets in chronological order; and a training unit that trains the model to learn the features of the learning data included in the set obtained by the division by the dividing unit, for each of the divided sets, in a predetermined order.

Inventors

  • Shinichiro Okamoto

Assignees

  • Actapio, Inc.

Dates

Publication Date
20260505
Application Date
20210909

Claims (8)

  1. 1 . A learning apparatus comprising: a processor, the processor is configured to: receive training data for learning model training; divide the training data into a plurality of data sets in chronological order; for each of a plurality of epochs: generate a plurality of combinations of subsets of data sets, wherein data sets in each subset are selected based on a predetermined order in each epoch; perform, for each of the plurality of combinations of subsets of data sets: connect the subset of data sets based on the predetermined order to generate a training data set; train a model instance to learn features of the training data set; and evaluate the model instance to determine model accuracy; and select a model instance with a highest model accuracy for deployment from the model instances of the plurality of epochs.
  2. 2 . The learning apparatus according to claim 1 , wherein the predetermined order is a random order.
  3. 3 . The learning apparatus according to claim 1 , wherein the predetermined order is in time series.
  4. 4 . The learning apparatus according to claim 1 , wherein the training data is divided into a number of data sets designated by a user.
  5. 5 . The learning apparatus according to claim 1 , wherein the training data is divided into a number of data sets that falls within a range designated by a user.
  6. 6 . A learning method to be executed by a learning apparatus, the method comprising: receiving, by a processor, training data for learning model training; dividing, by the processor, the training data into a plurality of data sets in chronological order; for each of a plurality of epochs: generating, by the processor a plurality of combinations of subsets of data sets, wherein data sets in each subset are selected based on a predetermined order in each epoch; performing, by the processor, for each of the plurality of combinations of subsets of data sets: connecting the subset of data sets based on the predetermined order to generate a training data set; training a model instance to learn features of the training data set; and evaluating the model instance to determine model accuracy; and selecting, by the processor, a model instance with a highest model accuracy for deployment from the model instances of the plurality of epochs.
  7. 7 . A non-transitory computer-readable storage medium having stored therein a learning program for causing a computer to execute: receiving training data for learning model training; dividing the training data into a plurality of data sets in chronological order; for each of a plurality of epochs: generating a plurality of combinations of subsets of data sets, wherein data sets in each subset are selected based on a predetermined order in each epoch; performing for each of the plurality of combinations of subsets of data sets: connecting the subset of data sets based on the predetermined order to generate a training data set; training a model instance to learn features of the training data set; and evaluating the model instance to determine model accuracy; and selecting a model instance with a highest model accuracy for deployment from the model instances of the plurality of epochs.
  8. 8 . A learning apparatus comprising: a processor, the processor is configured to: receive training data for learning model training; divide the training data into a plurality of data sets in chronological order; select a subset of data sets from the plurality of data sets in a predetermined order; connect the subset of data sets based on selection order to generate a training data group; divide the training data group into a plurality of training data sets according to a buffer size; shuffle the plurality of training data sets in a random order; and iteratively train the learning model by sequentially learning features of the plurality of training data sets in the random order.

Description

TECHNICAL FIELD The present invention relates to a learning apparatus, a learning method, and a learning program. BACKGROUND ART In recent years, there has been a proposed technique of training various models such as a support vector machine (SVM) and a deep neural network (DNN) to learn the features of learning data so that the model will perform various predictions and classifications. As an example of such a training method, there is a proposed technique of dynamically changing the learning mode of learning data in accordance with a hyperparameter value or the like. CITATION LIST Patent Literature Patent Literature 1: Patent Application Laid-Open No. 2019-164793 SUMMARY Technical Problem Unfortunately, however, it is difficult to ensure improvement of accuracy of the model with the above-described conventional technique. For example, in the above-described conventional technique, the learning data as a learning target of features is merely dynamically changed according to the values of the hyperparameter or the like. Therefore, when the hyperparameter values are not appropriate, there might be a case where improvement of the accuracy of the model fails. The present application has been made in view of the above, and aims to provide a learning apparatus, a learning method, and a non-transitory computer-readable storage medium having stored therein a learning program capable of improving the accuracy of a model. Solution to Problem It is an object of the present invention to at least partially solve the problems in the conventional technology. According to one aspect of an embodiment, A learning apparatus includes a dividing unit that divides predetermined learning data features of which are to be learned by a model by training, into a plurality of sets in chronological order. The learning apparatus includes a training unit that trains the model to learn the features of the learning data included in the set obtained by division by the dividing unit, for each of the divided sets, in a predetermined order. The above and other objects, features, advantages and technical and industrial significance of this invention will be better understood by reading the following detailed description of presently preferred embodiments of the invention, when considered in connection with the accompanying drawings. Advantageous Effects of Invention According to one aspect of the embodiment, there is an effect that accuracy of the model can be improved. BRIEF DESCRIPTION OF DRAWINGS FIG. 1 is a diagram illustrating an example of processing executed by an information providing device according to an embodiment; FIG. 2 is a diagram illustrating an example of an information processing system according to the embodiment; FIG. 3 is a diagram illustrating an overall picture of processes executed by an information processing device according to the embodiment; FIG. 4 is a diagram illustrating an example of division for each of trials when a data set is divided for each of applications; FIG. 5 is a diagram illustrating a configuration example of the information processing device according to the embodiment; FIG. 6 is a diagram conceptually illustrating the division of a data set; FIG. 7 is a diagram (1) illustrating a change in model performance when first and fourth optimization algorithms are executed; FIG. 8 is a diagram (2) illustrating a change in model performance when the first and fourth optimization algorithms are executed; FIG. 9 is a diagram illustrating a comparative example comparing the performance of models according to the combination of the first and fourth optimization algorithms; FIG. 10 is a diagram illustrating an example of a second optimization algorithm; FIG. 11 is a diagram illustrating an example of a third optimization algorithm; FIG. 12 is a diagram illustrating a comparative example in which the performance of the model is compared for individual shuffle buffer sizes; FIG. 13 is a diagram illustrating an example of conditional information regarding a fifth optimization algorithm; FIG. 14 is a diagram illustrating an example of the fifth optimization algorithm; FIG. 15 is a diagram illustrating an example of an optimization algorithm for optimizing a mask target; FIG. 16 is a diagram illustrating a comparative example in which the accuracy of the model is compared between a case where a mask target optimization is executed and a case where the mask target optimization is not executed; FIG. 17 is a diagram illustrating a configuration example of an execution control apparatus according to the embodiment; FIG. 18 illustrates an example of a model architecture storage unit according to the embodiment; FIG. 19 is a diagram illustrating an example of a model architecture associated with information indicating an execution target arithmetic unit; FIG. 20 is a diagram illustrating a state of performance improvement by experiments using a model for multi-class classification; FIG. 21 is a diagram illustrating an ex