Semi-supervised Regularization Research Articles

Learning-Based Regularization for Cardiac Strain Analysis via Domain Adaptation.

Reliable motion estimation and strain analysis using 3D+ time echocardiography (4DE) for localization and characterization of myocardial injury is valuable for early detection and targeted interventions. However, motion estimation is difficult due to the low-SNR that stems from the inherent image properties of 4DE, and intelligent regularization is critical for producing reliable motion estimates. In this work, we incorporated the notion of domain adaptation into a supervised neural network regularization framework. We first propose a semi-supervised Multi-Layered Perceptron (MLP) network with biomechanical constraints for learning a latent representation that is shown to have more physiologically plausible displacements. We extended this framework to include a supervised loss term on synthetic data and showed the effects of biomechanical constraints on the network’s ability for domain adaptation. We validated the semi-supervised regularization method on in vivo data with implanted sonomicrometers. Finally, we showed the ability of our semi-supervised learning regularization approach to identify infarct regions using estimated regional strain maps with good agreement to manually traced infarct regions from postmortem excised hearts.

Online Semi-Supervised Learning With Multiple Regularization Terms

Online semi-supervised learning (OS 2 L) has received much attention recently because of its well practical usefulness. Most of the existing studies of OS 2 L are related to manifold regularization. In this paper, we introduce a novel OS 2 L framework with multiple regularization terms based on the notion of ascending the dual function in constrained optimization. Using the Fenchel conjugate, different semi-supervised regularization terms can be integrated into the dual function easily and directly. This approach is derived by updating limited dual coefficient variables on each learning round. To be practical, we also employ buffering strategies and sparse approximation approaches in this paper. The experimental studies show that our methods achieve accuracy comparable to offline algorithms while consuming less time and memory. Especially, our OS 2 L algorithms can handle the settings where the target hyperplane of classification continually drifts with the sequence of arriving instances. This paper paves a way to design and analyze OS 2 L algorithms with multiple regularization terms.

Cross-media retrieval based on semi-supervised regularization and correlation learning

As large scale multimedia data in heterogeneous spaces is flooding into the Internet, cross-media retrieval is becoming increasingly significant. In cross-media retrieval, users can retrieve the results containing various types of media by submitting a query of any media type. However, most existing cross-media retrieval methods are restricted to the retrieval between two types of media, which ignores the semantic consistency of different media data. In addition, although some methods consider the similarity between same semantic category data in different media, they neglect the dissimilarity between different semantic category data in different media. To solve the above problems, we propose a novel feature learning algorithm for cross-media retrieval, called semi-supervised regularization and correlation learning (SSRCL), which is capable of modeling multiple types of media simultaneously. More importantly, SSRCL considers both semantic category similarity and dissimilarity simultaneously, and utilizes both labeled and unlabeled data to learn the projection matrices for different media types. The experimental results show that our proposed approach, compared with four state-of-the-art methods, has better performance on two extensively used datasets.

Cross-Media Retrieval Based on Query Modality and Semi-Supervised Regularization

Existing cross-media retrieval methods usually learn one same latent subspace for different retrieval tasks, which can only achieve a suboptimal retrieval. In this paper, we propose a novel cross-media retrieval method based on Query Modality and Semi-supervised Regularization (QMSR). Taking the cross-media retrieval between images and texts for example, QMSR learns two couples of mappings for different retrieval tasks (i.e. using images to search texts (Im2Te) or using texts to search images (Te2Im)) instead of learning one couple of mappings. QMSR learns two couples of projections by optimizing the correlation between images and texts and the semantic information of query modality (image or text), and integrates together the semi-supervised regularization, the structural information among both labeled and unlabeled data of query modality to transform different media objects from original feature spaces into two different isomorphic subspaces (Im2Te common subspace and Te2Im common subspace). Experimental results show the effectiveness of the proposed method.

Learning Cross-Media Joint Representation With Sparse and Semisupervised Regularization

Cross-media retrieval has become a key problem in both research and application, in which users can search results across all of the media types (text, image, audio, video, and 3-D) by submitting a query of any media type. How to measure the content similarity among different media is the key challenge. Existing cross-media retrieval methods usually focus on modeling the pairwise correlation or semantic information separately. In fact, these two kinds of information are complementary to each other and optimizing them simultaneously can further improve the accuracy. In this paper, we propose a novel feature learning algorithm for cross-media data, called joint representation learning (JRL), which is able to explore jointly the correlation and semantic information in a unified optimization framework. JRL integrates the sparse and semisupervised regularization for different media types into one unified optimization problem, while existing feature learning methods generally focus on a single media type. On one hand, JRL learns sparse projection matrix for different media simultaneously, so different media can align with each other, which is robust to the noise. On the other hand, both the labeled data and unlabeled data of different media types are explored. Unlabeled examples of different media types increase the diversity of training data and boost the performance of joint representation learning. Furthermore, JRL can not only reduce the dimension of the original features, but also incorporate the cross-media correlation into the final representation, which further improves the performance of both cross-media retrieval and single-media retrieval. Experiments on two datasets with up to five media types show the effectiveness of our proposed approach, as compared with the state-of-the-art methods.

Multitask Extreme Learning Machine for Visual Tracking

In this paper, we try to address the joint optimization problem of the extreme learning machines corresponding to different features. The method is based on the L 2,1 norm penalty, which encourages joint sparse coding. By adopting such a technology, the intrinsic relation between different features can be sufficiently preserved. To tackle the problem that the labeled samples is rare, we introduce the semi-supervised regularization term and seamlessly incorporate them into the particle filter framework to realize visual tracking. In addition, an online updating strategy is introduced which also exploits the large amount of unlabeled samples that are collected during the tracking period. Finally, the proposed tracking algorithm is compared to other state-of-the-arts on some challenging video sequences and shows promising results.