α-Geodesical Skew Divergence

Kimura, Masanari; Hino, Hideitsu

doi:10.3390/e23050528

Cited by 5 publications

(4 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To address this problem, we can utilize the idea of αgeodesical skew divergence [31]- [33], which is the generalization of KL-divergence.…”

Section: Generalization Bounds For the Decomposed Shiftsmentioning

confidence: 99%

On the Decomposition of Covariate Shift Assumption for the Set-to-Set Matching

Kimura

2023

IEEE Access

Self Cite

View full text Add to dashboard Cite

The task of set matching, which models the quality of matching between pairs of sets, is expected to have a wide range of practical applications. However, many existing methods that address this task assume that the training and testing distributions are identical, which is frequently violated in realworld scenarios. To address this issue, the covariate shift assumption focuses on the shift in the distribution of covariates between the training and testing datasets. While several studies have analyzed this assumption for vector inputs, there is a lack of research on similar assumptions when the input is a pair of sets. In this study, we refine and redefine the covariate shift assumption in set matching and analyze how models perform under these conditions.

show abstract

“…To address this problem, we can utilize the idea of αgeodesical skew divergence [31]- [33], which is the generalization of KL-divergence.…”

Section: Generalization Bounds For the Decomposed Shiftsmentioning

confidence: 99%

On the Decomposition of Covariate Shift Assumption for the Set-to-Set Matching

Kimura

2023

IEEE Access

Self Cite

View full text Add to dashboard Cite

show abstract

“…Future research direction include conducting quantitative evaluation experiments for abstract and difficult-to-evaluate problems, such as those covered in this study, creating datasets that enable these experiments (e.g., inspired by [48]), examining models that assume probability distributions other than the multidimensional Gaussian distribution, and further analysis on the best distance measure to use [49,50,51]. In particular, the framework of information geometry [37,37,52], which considers Riemannian manifolds formed by probability distributions, is very useful, and many machine learning algorithms have been analyzed [53,54,55,56,57,58].…”

Section: Limitationsmentioning

confidence: 99%

Fashion-Specific Attributes Interpretation via Dual Gaussian Visual-Semantic Embedding

Shimizu¹,

Kimura²,

Goto³

2022

Preprint

View full text Add to dashboard Cite

Several techniques to map various types of components, such as words, attributes, and images, into the embedded space have been studied. Most of them estimate the embedded representation of target entity as a point in the projective space. Some models, such as Word2Gauss, assume a probability distribution behind the embedded representation, which enables the spread or variance of the meaning of embedded target components to be captured and considered in more detail. We examine the method of estimating embedded representations as probability distributions for the interpretation of fashion-specific abstract and difficult-to-understand terms. Terms, such as "casual," "adult-casual," "beauty-casual," and "formal," are extremely subjective and abstract and are difficult for both experts and non-experts to understand, which discourages users from trying new fashion. We propose an end-to-end model called dual Gaussian visual-semantic embedding, which maps images and attributes in the same projective space and enables the interpretation of the meaning of these terms by its broad applications. We demonstrate the effectiveness of the proposed method through multifaceted experiments involving image and attribute mapping, image retrieval and re-ordering techniques, and a detailed theoretical/analytical discussion of the distance measure included in the loss function.

show abstract

“…Definition 4.1 (f -interpolation (Kimura and Hino, 2021)) For any a, b, ∈ R, some λ ∈ [0, 1] and some α ∈ R, we define f -interpolation as…”

Section: Statistical Model and Exponential Familymentioning

confidence: 99%

Information Geometrically Generalized Covariate Shift Adaptation

Kimura

Hino

2022

Neural Computation

Self Cite

View full text Add to dashboard Cite

Many machine learning methods assume that the training and test data follow the same distribution. However, in the real world, this assumption is often violated. In particular, the marginal distribution of the data changes, called covariate shift, is one of the most important research topics in machine learning. We show that the well-known family of covariate shift adaptation methods is unified in the framework of information geometry. Furthermore, we show that parameter search for a geometrically generalized covariate shift adaptation method can be achieved efficiently. Numerical experiments show that our generalization can achieve better performance than the existing methods it encompasses.

show abstract

α-Geodesical Skew Divergence

Cited by 5 publications

References 32 publications

On the Decomposition of Covariate Shift Assumption for the Set-to-Set Matching

On the Decomposition of Covariate Shift Assumption for the Set-to-Set Matching

Fashion-Specific Attributes Interpretation via Dual Gaussian Visual-Semantic Embedding

Information Geometrically Generalized Covariate Shift Adaptation

Contact Info

Product

Resources

About