US 12,299,067 B2
Sparse matrix standardization device, sparse matrix standardization method, sparse matrix standardization program, and data structure
Takuya Araki, Tokyo (JP)
Assigned to NEC CORPORATION, Tokyo (JP)
Appl. No. 17/298,291
Filed by NEC Corporation, Tokyo (JP)
PCT Filed Dec. 5, 2018, PCT No. PCT/JP2018/044765
§ 371(c)(1), (2) Date May 28, 2021,
PCT Pub. No. WO2020/115844, PCT Pub. Date Jun. 11, 2020.
Prior Publication US 2022/0050886 A1, Feb. 17, 2022
Int. Cl. G06F 17/16 (2006.01); G06F 7/499 (2006.01); G06F 17/18 (2006.01); G06N 3/08 (2023.01)
CPC G06F 17/16 (2013.01) [G06F 7/49936 (2013.01); G06F 17/18 (2013.01); G06N 3/08 (2013.01)] 8 Claims
OG exemplary drawing
 
1. A sparse matrix standardization device comprising:
a memory storing instructions; and
one or more processors configured to execute the instructions to:
perform, for each column of a sparse target matrix to be standardized and that is training data in machine learning, a computation process to compute an average and a standard deviation of a value of each component of the column;
perform, for each column of the sparse target matrix, a first dividing process to divide the value of each component of the column by the standard deviation computed on the basis of the column;
perform, for each column of the sparse target matrix, a second dividing process to divide the average computed on the basis of the column by the standard deviation computed on the basis of the column; and
arrange quotients computed by the second dividing process, in each row of the sparse target matrix in order of the columns from which the quotients are computed, to generate a row vector; and
a storage device configured to store, as the standardized sparse target matrix, a matrix computed by the first dividing process and the row vector generated for each row of the sparse target matrix.