Displaying iterative learning process and sample error to efficiently eliminate noisy sample data
According to one embodiment, a learning apparatus includes processing circuitry. The processing circuitry acquires a plurality of learning samples to be learned and a plurality of target labels associated with the respective learning samples, iteratively learns a learning model so that a learning error between output data corresponding to the learning sample and the target label is small with respect to the learning model to which the output data is output by inputting the learning sample, and displays a layout image in which at least some of the learning samples are arranged based on a learning progress regarding the iterative learning of the learning model and a plurality of the learning errors.
1 . A learning apparatus comprising processing circuitry configured to:
acquire a plurality of learning samples to be learned and a plurality of target labels associated with the respective learning samples;
iteratively cause learning by a learning model which receives an input of a learning sample, among the plurality of learning samples, and outputs output data and a learning error, the learning error representing an error between the output data corresponding to the inputted learning sample and a target label, among the plurality of target labels; and
display, based on a learning progress regarding the iterative learning of the learning model and a plurality of learning errors, a layout image in which at least some of the learning samples are arranged for each point in the learning progress and according to the learning errors.
2 . The apparatus according to claim 1 , wherein the learning progress is a number of times of the iterative learning.
3 . The apparatus according to claim 1 , wherein the at least some learning samples correspond to a plurality of representative values of the learning errors.
4 . The apparatus according to claim 1 , wherein the at least some learning samples correspond to a specific target label.
5 . The apparatus according to claim 1 , wherein the processing circuitry is further configured to arrange the at least some learning samples in a column direction in ascending or descending order of the learning error, and display the layout image in which the learning progress is associated in a row direction.
6 . The apparatus according to claim 5 , wherein the processing circuitry is further configured to perform emphasis processing based on the learning errors on the learning sample included in the layout image.
7 . The apparatus according to claim 6 , wherein the emphasis processing is at least one of a change in thickness of a frame line of the learning sample, among the plurality of learning samples, a change in color of the frame line of the learning sample, a superimposed display of a color map on the learning sample, and a change in display size of the learning sample.
8 . The apparatus according to claim 5 , wherein the processing circuitry is further configured to perform emphasis processing on a same learning sample on the layout image.
9 . The apparatus according to claim 1 , wherein the processing circuitry is further configured to display, as the layout image, an error map image that is associated in a column direction in ascending or descending order of the learning error and in which the learning progress is associated in a row direction.
10 . The apparatus according to claim 9 , wherein the processing circuitry is further configured to perform emphasis processing on a position corresponding to a same learning sample on the error map image.
11 . The apparatus according to claim 1 , wherein the processing circuitry is further configured to:
set a boundary line on the layout image; and
exclude one or more learning samples designated by the set boundary line from the learning samples to cause learning by the learning model.
12 . The apparatus according to claim 11 , wherein the processing circuitry is further configured to:
determine an arbitrary time point in the learning progress based on the set boundary line and
exclude the one or more learning samples to restart iterative learning from the arbitrary time point.
13 . The apparatus according to claim 11 , wherein the boundary line is at least one of a straight line, a curved line, and a plane figure.
14 . The apparatus according to claim 11 , wherein the processing circuitry is further configured to:
set a range on the layout image; and
further display details of one or more learning samples corresponding to the set range.
15 . The apparatus according to claim 1 , wherein the processing circuitry is further configured to:
set a range on the layout image; and
further display details of one or more learning samples corresponding to the set range.
16 . The apparatus according to claim 1 , wherein the learning model is a neural network model.
17 . A learning method comprising:
acquiring a plurality of learning samples to be learned and a plurality of target labels associated with the respective learning samples;
iteratively cause learning by a learning model which receives an input of a learning sample, among the plurality of learning sample, and outputs output data and a learning error, the learning error representing an error between the output data corresponding to the inputted learning sample and a target label, among the plurality of target labels; and
displaying, based on a learning progress regarding the iterative learning of the learning model and a plurality of learning errors, a layout image in which at least some of the learning samples are arranged for each point in the learning progress and according to the learning errors.
18 . A non-transitory computer-readable storage medium storing a program for causing a computer to execute processing comprising:
acquiring a plurality of learning samples to be learned and a plurality of target labels associated with the respective learning samples;
iteratively cause learning by a learning model which receives an input of a learning sample, among the plurality of learning sample, and outputs output data and a learning error, the learning error representing an error between the output data corresponding to the inputted learning sample and a target label, among the plurality of target labels; and
displaying, based on a learning progress regarding the iterative learning of the learning model and a plurality of learning errors, a layout image in which at least some of the learning samples are arranged for each point in the learning progress and according to the learning errors.