Information processing apparatus, information processing method, and storage medium
An information processing apparatus includes one or more memories, and one or more processors that, when executing instructions stored in the one or more memories, function as the following units: an acquisition unit configured to acquire learning data including data and a label indicating a category of the data, a base holding unit configured to hold a base for generating a representative vector in the category, a learning unit configured to learn a parameter related to generation of the representative vector based on the acquired learning data, and a first generation unit configured to generate the representative vector based on the parameter and the base.
1 . An information processing apparatus comprising: one or more processors; and one or more memories that store instructions for causing the one or more processors and the one or more memories to: acquire learning data including first data, second data, a label indicating a first category of the first data, and a label indicating a second category of the second data; hold a plurality of base vectors for generating a first representative vector for the first category and a second representative vector for the second category; learn a plurality of first parameters related to generation of the first representative vector based on the first data and a plurality of second parameters related to generation of the second representative vector based on the second data; generate the first representative vector based on the plurality of first parameters and the plurality of base vectors, and generate the second representative vector based on the plurality of second parameters and the plurality of base vectors; obtain a feature vector; and determine a likelihood that third data from which the feature vector was generated belongs to the first category based on the feature vector and on the first representative vector, wherein a total memory size of the plurality of base vectors is smaller than a total memory size of a plurality of representative vectors that include the first representative vector and the second representative vector, wherein each of the plurality of representative vectors can be generated from the plurality of base vectors.
2 . The information processing apparatus according to claim 1 ,
wherein the one or more memories further store instructions for causing the one or more processors and the one or more memories to generate a feature vector based on the first data included in the learning data, and
wherein the plurality of first parameters include a parameter related to the generation of the feature vector and the plurality of base vectors.
3 . The information processing apparatus according to claim 2 , wherein the one or more memories further store instructions for causing the one or more processors and the one or more memories to learn the plurality of first parameters related to the generation of the first representative vector using the learning data, the first representative vector, and the feature vector.
4 . The information processing apparatus according to claim 1 ,
wherein the one or more memories further store instructions for causing the one or more processors and the one or more memories to generate unit the first representative vector by linearly combining the plurality of base vectors.
5 . The information processing apparatus according to claim 4 ,
wherein the plurality of first parameters are respective a weights for linearly combining the plurality of base vectors, and
wherein the one or more memories further store instructions for causing the one or more processors and the one or more memories to generate the first representative vector by linearly combining the plurality of base vectors based on the weights for linearly combining the plurality of base vectors.
6 . The information processing apparatus according to claim 4 , wherein the plurality of base vectors include three or more base vectors.
7 . The information processing apparatus according to claim 1 ,
wherein the plurality of base vectors include a first base vector, and
wherein the one or more memories further store instructions for causing the one or more processors and the one or more memories to set the first base vector and a vector sign-inverted from the first base vector as the first representative vector.
8 . The information processing apparatus according to claim 7 , wherein the one or more memories further store instructions for causing the one or more processors and the one or more memories to assign two labels with low degrees of similarity to the first base vector based on an initial degree of similarity calculated based on the learning data.
9 . The information processing apparatus according to claim 1 ,
wherein the one or more memories further store instructions for causing the one or more processors and the one or more memories to set a first base vector, of the plurality of base vectors, and a vector in which at least some elements of the first base vector are sign-inverted, as the first representative vector.
10 . The information processing apparatus according to claim 1 ,
wherein the one or more memories further store instructions for causing the one or more processors and the one or more memories to set a first base vector, of the plurality of base vectors, and a vector in which at least some elements of the first base vector are interchanged, as the first representative vector.
11 . The information processing apparatus according to claim 1 ,
wherein the plurality of first parameters include weights for linearly combining the plurality of base vectors, and
wherein each weight of the weights corresponds to a respective base vector of the plurality of base vectors.
12 . The information processing apparatus according to claim 11 ,
wherein the one or more memories further store instructions for causing the one or more processors and the one or more memories to generate the first representative vector by linearly combining the plurality of base vectors based on the respective weight that corresponds to each base vector of the plurality of base vectors.
13 . The information processing apparatus according to claim 1 , wherein the one or more memories further store instructions for causing the one or more processors and the one or more memories to: determine a likelihood that the third data from which the feature vector was generated belongs to the second category based on the feature vector and on the second representative vector.
14 . The information processing apparatus according to claim 13 , wherein the first data includes the third data.
15 . The information processing apparatus according to claim 1 , wherein a number of the plurality of base vectors is smaller than a number of categories for which respective representative vectors are generated.
16 . An information processing method executed by an information processing apparatus, the information processing method comprising: acquiring learning data including first data, second data, a label indicating a first category of the first data, and a label indicating a second category of the second data; holding a plurality of base vectors for generating a first representative vector for the first category and a second representative vector for the second category; learning a plurality of first parameters related to generation of the first representative vector based on the first data and a plurality of second parameters related to generation of the second representative vector based on the second data; and generating the first representative vector based on the plurality of first parameters and the plurality of base vectors and generating the second representative vector based on the plurality of second parameters and the plurality of base vectors; obtaining a feature vector; and determining a likelihood that third data from which the feature vector was generated belongs to the first category based on the feature vector and on the first representative vector, wherein a total memory size of the plurality of base vectors is smaller than a total memory size of a plurality of representative vectors that include the first representative vector and the second representative vector, wherein each of the plurality of representative vectors can be generated from the plurality of base vectors.
17 . A non-transitory computer-readable storage medium that stores computer-executable instructions that, when executed by a computer, cause the computer to: acquire learning data including first data, second data, a label indicating a first category of the first data, and a label indicating a second category of the second data; hold a plurality of base vectors for generating a first representative vector for the first category and a second representative vector for the second category; learn a plurality of first parameters related to generation of the first representative vector based on the first data and a plurality of second parameters related to generation of the second representative vector based on the second data; and generate the first representative vector based on the plurality of first parameters and the plurality of base vectors and generate the second representative vector based on the plurality of second parameters and the plurality of base vectors; obtain a feature vector; and determine a likelihood that third data from which the feature vector was generated belongs to the first category based on the feature vector and on the first representative vector, wherein a total memory size of the plurality of base vectors is smaller than a total memory size of a plurality of representative vectors that include the first representative vector and the second representative vector, wherein each of the plurality of representative vectors can be generated from the plurality of base vectors.