Computer-readable recording medium storing learning support program, learning support method, and learning support device
A recording medium stores a program for causing a computer to execute a process including: calculating a first embedded vector for each cluster obtained by clustering samples in training data, by inputting the samples that represent the clusters to a first distance metric model; performing training of a second distance metric model, based on labels set in pairs of the samples; calculating a second embedded vector for each cluster, by inputting the samples that represent the clusters to the second distance metric model; detecting pairs of the clusters that are likely to be integrated when the training is performed with a greater number of epochs than a number of epochs at a time of the training of the second distance metric model, based on the first embedded vector and the second embedded vector; and outputting one of the pairs of the clusters in which a similarity label is not set.
1 . A non-transitory computer-readable recording medium storing a training support program for causing a computer to execute a process comprising:
calculating a first embedded vector for each of clusters obtained by clustering samples included in training data, by inputting the samples that represent the clusters to a first distance metric model;
performing training of a second distance metric model from the first distance metric model, based on labels set in pairs of the samples included in the training data, the training being a provisional training performed for a predetermined, limited number of epochs, substantially fewer than a number of epochs required for model convergence;
calculating a second embedded vector for each of the clusters, by inputting the samples that represent the clusters to the provisionally trained second distance metric model;
detecting pairs of the clusters that are likely to be integrated when the training is performed with a greater number of epochs than the number of epochs, based on comparison of the first embedded vector and the second embedded vector, the detecting including calculating an amount of change in a distance between the clusters before and after the provisional training;
identifying, from among the detected pairs of the clusters, at least one pair being a candidate for unintended integration in which a similarity label is not set; and
generating, on a graphical user interface, an interactive alert for controlling the retraining process, the alert notifying a developer of the at least one pair and comprising an option to select one of modifying training labels or interrupting retraining process.
2 . The non-transitory computer-readable recording medium according to claim 1 , wherein the outputting includes outputting a recommendation formed to set a dissimilarity label to the one of the pairs of the clusters in which the similarity label is not set.
3 . A training support method comprising:
calculating a first embedded vector for each of clusters obtained by clustering samples included in training data, by inputting the samples that represent the clusters to a first distance metric model;
performing training of a second distance metric model from the first distance metric model, based on labels set in pairs of the samples included in the training data, the training being a provisional training performed for a predetermined, limited number of epochs, substantially fewer than a number of epochs required for model convergence;
calculating a second embedded vector for each of the clusters, by inputting the samples that represent the clusters to the provisionally trained second distance metric model;
detecting pairs of the clusters that are likely to be integrated when the training is performed with a greater number of epochs than the number of epochs, based on comparison of the first embedded vector and the second embedded vector, the detecting including calculating an amount of change in a distance between the clusters before and after the provisional training;
identifying, from among the detected pairs of the clusters, at least one pair being a candidate for unintended integration in which a similarity label is not set; and
generating, on a graphical user interface, an interactive alert for controlling the retraining process, the alert notifying a developer of the at least one pair and comprising an option to select one of modifying training labels or interrupting retraining process.
4 . The training support method according to claim 3 , wherein the outputting includes outputting a recommendation formed to set a dissimilarity label to the one of the pairs of the clusters in which the similarity label is not set.
5 . An information processing device comprising:
a memory; and
a processor coupled to the memory and configured to:
calculate a first embedded vector for each of clusters obtained by clustering samples included in training data, by inputting the samples that represent the clusters to a first distance metric model;
perform training of a second distance metric model from the first distance metric model, based on labels set in pairs of the samples included in the training data, the training being a provisional training performed for a predetermined, limited number of epochs, substantially fewer than a number of epochs required for model convergence;
calculate a second embedded vector for each of the clusters, by inputting the samples that represent the clusters to the provisionally trained second distance metric model;
detect pairs of the clusters that are likely to be integrated when the training is performed with a greater number of epochs than the number of epochs, based on comparison of the first embedded vector and the second embedded vector, the detecting including calculating an amount of change in a distance between the clusters before and after the provisional trainingd;
identify, from among the detected pairs of the clusters, at least one pair being a candidate for unintended integration in which a similarity label is not set; and
generate, on a graphical user interface, an interactive alert for controlling the retraining process, the alert notifying a developer of the at least one pair and comprising an option to select one of modifying training labels or interrupting retraining process.
6 . The information processing device according to claim 5 , wherein the processor outputs a recommendation formed to set a dissimilarity label to the one of the pairs of the clusters in which the similarity label is not set.