Information processing apparatus, information processing method, and storage medium
Before dimension reduction is performed while local data distribution is stored as neighborhood data, a distance between data to be subjected to the dimension reduction is calculated, and a parameter (a neighborhood number of the k-nearest neighbor algorithm or a size of a hypersphere) which determines the neighborhood data is determined for each data to be subjected to the dimension reduction. Thereafter, the dimension reduction is performed on the target data based on the determined parameter.
1. An apparatus which performs dimension reduction of a plurality of high dimensional data each representing a plurality of image features extracted from images, the apparatus comprising:
one or more processors; and
one or more memories coupled to the one or more processors, the memories stored thereon instructions which, when executed by the one or more processors, cause the apparatus to:
calculate a distance between each pair of high dimensional data in the plurality of high dimensional data;
determine a plurality of parameters each of which determines neighborhood data for respective one of the plurality of high dimensional data based on data distribution with respect to each of the plurality of high dimensional data using the calculated distance between each pair of the plurality of high dimensional data;
perform dimension reduction on the plurality of high dimensional data while keeping the neighborhood data for each of the plurality of high dimensional data in a neighborhood based on respective determined parameters;
input information on a distance relationship of at least part of data to be subjected to the dimension reduction;
re-calculate a parameter which determines the neighborhood data based on the input information; and
perform the dimension reduction on the data to be subjected to the dimension reduction based on the re-calculated parameter.
2. The apparatus according to claim 1 ,
wherein different types of data to be subjected to the dimension reduction have different labels, and
wherein the instruction further cause the apparatus to determine the parameter which determines the neighborhood data based on distribution of data having a same label for each data to be subjected to the dimension reduction.
3. The apparatus according to claim 2 , wherein the instruction further cause the apparatus to determine the parameter which determines the neighborhood data such that a same parameter is shared by data having the same label based on the distribution of the data having the same label.
4. The apparatus according to claim 3 , wherein the instructions further cause the apparatus to:
input an initial value of the parameter which determines the neighborhood data, and
determine the parameter which determines the neighborhood data by searching for the parameter from the initial value.
5. The apparatus according to claim 1 wherein the instructions further cause the apparatus to update the neighborhood data by calculating the parameter which determines the neighborhood data for each data to be subjected to the dimension reduction.
6. The apparatus according to claim 1 , wherein the parameter which determines the neighborhood data is a neighborhood number of a k-nearest neighbor algorithm or a size of a hypersphere.
7. A method for performing dimension reduction of a plurality of high dimensional data each representing a plurality of image features extracted from images, the method comprising:
calculating a distance between each pair of high dimensional data in the plurality of high dimensional data;
determining a plurality of parameters each of which determines neighborhood data for respective one of the plurality of high dimensional data based on data distribution with respect to each of the plurality of high dimensional data using the calculated distance between each pair of the plurality of high dimensional data;
performing dimension reduction on the plurality of high dimensional data while keeping the neighborhood data for each of the plurality of high dimensional data in a neighborhood based on respective determined parameters;
inputting information on a distance relationship of at least part of data to be subjected to the dimension reduction;
re-calculating a parameter which determines the neighborhood data based on the input information; and
performing the dimension reduction on the data to be subjected to the dimension reduction based on the re-calculated parameter.
8. The method according to claim 7 ,
wherein different types of data to be subjected to the dimension reduction have different labels, and
determining determines the parameter which determines the neighborhood data based on distribution of data having a same label for each data to be subjected to the dimension reduction.
9. The method according to claim 8 , wherein the determining determines the parameter which determines the neighborhood data such that a same parameter is shared by data having the same label based on the distribution of the data having the same label.
10. The method according to claim 9 , further comprising:
inputting an initial value of the parameter which determines the neighborhood data,
wherein the determining determines the parameter which determines the neighborhood data by searching for the parameter from the initial value.
11. The method according to claim 7 wherein the determining updates the neighborhood data by calculating the parameter which determines the neighborhood data for each data to be subjected to the dimension reduction.
12. The method according to claim 7 , wherein the parameter which determines the neighborhood data is a neighborhood number of a k-nearest neighbor algorithm or a size of a hypersphere.
13. A non-transitory storage medium storing a program which controls an information processing apparatus which performs dimension reduction of a plurality of high dimensional data each representing a plurality of image features extracted from images, the program causing a computer to execute:
calculating a distance between each pair of high dimensional data in the plurality of high dimensional data;
determining a plurality of parameters each of which determines neighborhood data for respective one of the plurality of high dimensional data based on data distribution with respect to each of the plurality of high dimensional data using the calculated distance between each pair of the plurality of high dimensional data;
performing dimension reduction on the plurality of high dimensional data while keeping the neighborhood data for each of the plurality of high dimensional data in a neighborhood based on respective determined parameters;
inputting information on a distance relationship of at least part of data to be subjected to the dimension reduction;
re-calculating a parameter which determines the neighborhood data based on the input information; and
performing the dimension reduction on the data to be subjected to the dimension reduction based on the re-calculated parameter.
14. The non-transitory storage medium according to claim 13 ,
wherein different types of data to be subjected to the dimension reduction have different labels, and
determining determines the parameter which determines the neighborhood data based on distribution of data having a same label for each data to be subjected to the dimension reduction.
15. The non-transitory storage medium according to claim 14 , wherein the determining determines the parameter which determines the neighborhood data such that a same parameter is shared by data having the same label based on the distribution of the data having the same label.
16. The non-transitory storage medium according to claim 13 wherein the determining updates the neighborhood data by calculating the parameter which determines the neighborhood data for each data to be subjected to the dimension reduction.
17. The non-transitory storage medium according to claim 13 , wherein the parameter which determines the neighborhood data is a neighborhood number of a k-nearest neighbor algorithm or a size of a hypersphere.