US 8,805,757 B2Granted

Evaluation predicting device, evaluation predicting method, and program

AssigneeSony Corporation

InventorMasashi Sekino

Life of this patent

Estimating the term…

Loading patent…

Abstract

Disclosed herein is an evaluation predicting device including: an estimating section configured to define a plurality of first latent vectors, a plurality of second latent vectors, evaluation values, a plurality of first feature vectors, a plurality of second feature vectors, a first projection matrix, and a second projection matrix, express the first latent vectors and the second latent vectors, and perform Bayesian estimation with the first feature vectors, the second feature vectors, and a known the evaluation value as learning data, and calculate a posterior distribution of a parameter group including the first latent vectors, the second latent vectors, the first projection matrix, and the second projection matrix; and a predicting section configured to calculate a distribution of an unknown the evaluation value on a basis of the posterior distribution of the parameter group.

Claims11

Dependent claims

A hardware configuration of an information processing device comprising a microprocessor configured to execute evaluation prediction functions, the microprocessor comprising:

an estimating section configured to define
a plurality of first latent vectors indicating features latently possessed by a plurality of first items, respectively,
a plurality of second latent vectors indicating features latently possessed by a plurality of second items, respectively,
evaluation values corresponding to respective combinations of said first items and said second items and expressed by inner products of said first latent vectors and said second latent vectors,
a plurality of first feature vectors indicating known features possessed by the plurality of said first items,
a plurality of second feature vectors indicating known features possessed by the plurality of said second items,
a first projection matrix for projecting said first feature vectors into a space of said first latent vectors, and
a second projection matrix for projecting said second feature vectors into a space of said second latent vectors,
express said first latent vectors by a normal distribution having projection values of said first feature vectors projected by said first projection matrix as expected values, and express said second latent vectors by a normal distribution having projection values of said second feature vectors projected by said second projection matrix as expected values, and
perform Bayesian estimation with said first feature vectors, said second feature vectors, and a known said evaluation value as learning data, and calculate a posterior distribution of a parameter group including said first latent vectors, said second latent vectors, said first projection matrix, and said second projection matrix;
and
a predicting section configured to calculate a distribution of an unknown said evaluation value on a basis of the posterior distribution of said parameter group and calculate an expected value of said unknown evaluation value on the basis of the posterior distribution of said parameter group.

9 more elements

A hardware configuration of an information processing device comprising a microprocessor configured to execute evaluation prediction functions, the microprocessor comprising:

an estimating section configured to define
N first latent vectors ui(t) (i=1, . . . , N) indicating features latently possessed by N first items, respectively, at time t,
M second latent vectors vj(t) (j=1, . . . , M) indicating features latently possessed by M second items, respectively, at time t,
evaluation values yij(t) corresponding to respective combinations of said first items and said second items and expressed by inner products of said first latent vectors ui(t) and said second latent vectors vj(t) at time t,
a first projection matrix for projecting first latent vectors ui(t−1) at time (t−1) into a space of said first latent vectors ui(t) at time t, and
a second projection matrix for projecting second latent vectors vj(t−1) at time (t−1) into a space of said second latent vectors vj(t) at time t,
express the first latent vectors ui(t) at time t by a normal distribution having projection values obtained by projecting the first latent vectors ui(t−1) at time (t−1) by said first projection matrix as expected values, and express the second latent vectors vj(t) at time t by a normal distribution having projection values obtained by projecting the second latent vectors vj(t−1) at time (t−1) by said second projection matrix as expected values, and
perform Bayesian estimation with the first latent vectors ui(t−1), the second latent vectors vj(t−1), and evaluation values yij(t−1) at time (t−1) as learning data, and calculate a posterior distribution of a parameter group including the first latent vectors ui(t), the second latent vectors vj(t), said first projection matrix, and said second projection matrix at time t;
and
a predicting section configured to calculate an expected value of an evaluation value yij(t) at time t on a basis of the posterior distribution of said parameter group at time t.

7 more elements

A non-transitory, computer-readable medium comprising instructions for making a computer realize:

an estimating function of defining
a plurality of first latent vectors indicating features latently possessed by a plurality of first items, respectively,
a plurality of second latent vectors indicating features latently possessed by a plurality of second items, respectively,
evaluation values corresponding to respective combinations of said first items and said second items and expressed by inner products of said first latent vectors and said second latent vectors,
a plurality of first feature vectors indicating known features possessed by the plurality of said first items,
a plurality of second feature vectors indicating known features possessed by the plurality of said second items,
a first projection matrix for projecting said first feature vectors into a space of said first latent vectors, and
a second projection matrix for projecting said second feature vectors into a space of said second latent vectors,
expressing said first latent vectors by a normal distribution having projection values of said first feature vectors projected by said first projection matrix as expected values, and expressing said second latent vectors by a normal distribution having projection values of said second feature vectors projected by said second projection matrix as expected values, and
performing Bayesian estimation with said first feature vectors, said second feature vectors, and a known said evaluation value as learning data, and calculating a posterior distribution of a parameter group including said first latent vectors, said second latent vectors, said first projection matrix, and said second projection matrix;
and
a predicting function of calculating a distribution of an unknown said evaluation value on a basis of the posterior distribution of said parameter group and calculating an expected value of said unknown evaluation value on the basis of the posterior distribution of said parameter group.

9 more elements

Dependencies are inferred from the claim text.