Information processing apparatus, information processing method, imaging device, vehicle device, and medical robot device
There is provided an information processing apparatus for artificially increasing data of a minority attribute to generate learning data for making fair determination on each piece of input data. The information processing apparatus includes a data holding unit configured to hold first learning data to be used for learning a machine learning model an acquisition unit configured to acquire information regarding bias of the learning data a data generation unit configured to generate second learning data by using data included in the learning data on the basis of the information regarding bias and a learning unit configured to learn the machine learning model by using the first learning data and the second learning data.
1 . An information processing apparatus, comprising:
circuitry configured to:
hold first learning data for a learning operation of a machine learning model;
acquire information regarding bias of the first learning data, wherein
the acquired information regarding the bias corresponds to a set of data associated with at least one specific attribute, and
the first learning data includes the set of data;
superimpose noise on the first learning data;
generate second learning data based on the superimposed noise on the first learning data and the information regarding the bias of the first learning data; and
execute, based on the generated second learning data and the held first learning data, the learning operation of the machine learning model.
2 . The information processing apparatus according to claim 1 , wherein the circuitry is further configured to:
acquire information indicating a minority attribute in the held first learning data; and
generate, further based on the information indicating the minority attribute, the second learning data, wherein the generated second learning data have a same attribute as the minority attribute.
3 . The information processing apparatus according to claim 2 , wherein
the circuitry is further configured to generate, based on the minority attribute, Adversarial Example as the second learning data.
4 . The information processing apparatus according to claim 3 , wherein
the circuitry is further configured to generate the Adversarial Example further based on a Fast Gradient Sign Method.
5 . An information processing method, comprising:
storing first learning data for a learning operation of a machine learning model;
acquiring information regarding bias of the first learning data, wherein
the acquired information regarding the bias corresponds to a set of data associated with at least one specific attribute, and
the first learning data includes the set of data;
superimposing noise on the first learning data;
generating second learning data based on the superimposed noise on the first learning data and the information regarding the bias of the first learning data; and
executing, based on the generated second learning data and the stored first learning data, the learning operation of the machine learning model.
6 . A non-transitory computer-readable medium having stored thereon, computer-executable instructions which, when executed by a computer, cause the computer to execute operations, the operations comprising:
holding first learning data for a learning operation of a machine learning model;
acquiring information regarding bias of the first learning data, wherein
the acquired information regarding the bias corresponds to a set of data associated with at least one specific attribute, and
the first learning data includes the set of data;
superimposing noise on the first learning data;
generating second learning data based on the superimposed noise on the first learning data and the information regarding the bias of the first learning data; and
executing, based on the generated second learning data and the held first learning data, the learning operation of the machine learning model.
7 . An imaging device, comprising:
circuitry configured to:
hold first learning data for a learning operation of a machine learning model;
acquire information regarding bias of the first learning data, wherein
the acquired information regarding the bias corresponds to a set of data associated with at least one specific attribute, and
the first learning data includes the set of data;
superimpose noise on the first learning data;
generate second learning data based on the superimposed noise on the first learning data and the information regarding the bias of the first learning data;
execute, based on the generated second learning data and the held first learning data, the learning operation of the machine learning model;
capture an image; and
recognize the captured image based on the machine learning model and the executed learning operation of the machine learning model.
8 . The imaging device according to claim 7 , wherein the circuitry is further configured to:
determine, based on a field associated with the imaging device, a minority in the first learning data; and
generate the second learning data based on data of an image associated with the determined minority in the first learning data.
9 . A vehicle device, comprising:
circuitry configured to:
hold first learning data for a learning operation of a machine learning model;
acquire information regarding bias of the first learning data, wherein
the acquired information regarding the bias corresponds to a set of data associated with at least one specific attribute, and
the first learning data includes the set of data;
superimpose noise on the first learning data;
generate second learning data based on the superimposed noise on the first learning data and the information regarding the bias of the first learning data;
execute, based on the generated second learning data and the held first learning data, the learning operation of the machine learning model;
capture an image in a vicinity of the vehicle device; and
recognize the captured image based on the machine learning model and
the executed learning operation of the machine learning model.
10 . A medical robot device, comprising:
circuitry configured to:
hold first learning data for a learning operation of a machine learning model;
acquire information regarding bias of the first learning data, wherein
the acquired information regarding the bias corresponds to a set of data associated with at least one specific attribute, and
the first learning data includes the set of data;
superimpose noise on the first learning data;
generate second learning data based on the superimposed noise on the first learning data and the information regarding the bias of the first learning data;
execute, based on the generated second learning data and the held first learning data, the learning operation of the machine learning model;
capture an image in a vicinity of a surgical site; and
recognize the captured image based on the machine learning model and the executed learning operation of the machine learning model.