Domain adaption based on source dictionary regularized RKHS subspace learning

Lei, Wenjie; Ma, Zhengming; Lin, Yuanping; Gao, Wenxu

doi:10.1007/s10044-021-01002-x

Cited by 4 publications

(5 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Here, we compare our MSEpRKHS-DA model with seven methods on four real dataset to evaluate its performance. And these methods are TCA [12], SSTCA [12], IGLDA [14], TIT [16], CDSPP [20], SDRKHS-DA [19], and LPJT [18]. And the knn (k = 1) classifier is used.…”

Section: B Compare Mseprkhs-da Model With Some State-of-the-art Methodsmentioning

confidence: 99%

“…What's more, as in TIT, LPJT also utilizes the two different transformations for subspace learning, one for each domain, so LPJT can be applied in HDA too. For the SDRKHS-DA method [19], it also considers aligning the distributions by MMD criterion, in the meantime, introduces the dictionary learning into the RKHS subspace learning framework. That is, SDRKHS-DA uses source domain data as dictionary to code the target domain data and keeps the coding as sparse as possible at the same time, which makes the same kind of source domain data and target domain data close to each other in subspace to improve the performance of domain adaption.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Domain Adaption Based on MSE Criterion and Progressive RKHS Subspace Learning (MSEpRKHS_DA)

Qiu

Liu

et al. 2022

IEEE Access

View full text Add to dashboard Cite

Reproducing Kernel Hilbert Space (RKHS) subspace learning is very popular among the domain adaption, which learns a latent RKHS subspace for the source domain and target domain, so that their distribution gap becomes smaller than in the original data space. There is a famous probability theory: two second-order moment random variables are equal if and only if their mean squared error (MSE) is zero. In this paper, firstly, we use second-order moment random variables to model the source domain and target domain. Then, we prove that a second-order moment random variable is still second-order moment after it is transformed into the RKHS subspace. Finally, we propose the MSE criterion to measure the distribution difference between source domain and target domain. To our best knowledge, we are the first to apply the MSE to RKHS subspace learning. And the experiments show the superiority of MSE criterion, which performs better than the common Maximum Mean Difference (MMD) and the Covariance Matrix (CovM) criteria. Furthermore, considering the robustness of the RKHS subspace learning framework to the data dimension, we propose the domain adaption framework of the progressive RKHS subspace learning (pRKHS-DA), which continuously updates the learned RKHS subspace. Each update takes the previous learned subspace as the starting point. The idea of pRKHS-DA is proposed for the first time in this paper. Finally, this paper proposes MSEpRKHS_DA model based on MSE criterion and pRKHS-DA framework. And experiments show that our model achieves higher classification accuracy than some state-of-the-art methods.INDEX TERMS Domain adaption, MSE criterion, RKHS subspace learning. TABLE 7. The performance of our model and the baseline methods on 8 domain adaption tasks of ORL dataset.

show abstract

Section: B Compare Mseprkhs-da Model With Some State-of-the-art Methodsmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Domain Adaption Based on MSE Criterion and Progressive RKHS Subspace Learning (MSEpRKHS_DA)

Qiu

Liu

et al. 2022

IEEE Access

View full text Add to dashboard Cite

show abstract

“…, that is defined in (17), also grows as W grows and L increases from L i to L i+1 = L i + M . Matrix U is used to find vector h in (20) for SA and can be updated as (refer to appendix A.1):…”

Section: Adding Data Vectorsmentioning

confidence: 99%

“…where v T m = u T m W i Λ or if we prefer v m = ΛW T i u m . Note that v m is defined similar to equation (31), but elements m are set to zero and C i W i Λ is defined similar to U i in (17), with columns m equal to zero; thus Û equals U i + u m α m v T m , with columns m set to zero. Equation ( 38) can be rewritten as…”

Section: Removing Data Vectorsmentioning

confidence: 99%

“…To this end, data samples are first nonlinearly mapped into a higher (or even infinite) dimensional Reproducing Kernel Hilbert Space (RKHS), known as the feature space, and are then sparsely represented based on a dictionary in the same space. A few examples of the recent applications of the KSRC are medical applications [12,13], image quality assessment [14], image classification [15], computer vision [16], domain adaption [17] and fault diagnosis [18].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation