Liantao Wang

Bio: Liantao Wang is an academic researcher from Hohai University. The author has contributed to research in topics: Support vector machine & Discriminative model. The author has an hindex of 7, co-authored 21 publications receiving 175 citations. Previous affiliations of Liantao Wang include Carnegie Mellon University & Nanjing University of Science and Technology.

Multiple kernel clustering based on centered kernel alignment

Abstract: Multiple kernel clustering (MKC), which performs kernel-based data fusion for data clustering, is an emerging topic. It aims at solving clustering problems with multiple cues. Most MKC methods usually extend existing clustering methods with a multiple kernel learning (MKL) setting. In this paper, we propose a novel MKC method that is different from those popular approaches. Centered kernel alignment-an effective kernel evaluation measure-is employed in order to unify the two tasks of clustering and MKL into a single optimization framework. To solve the formulated optimization problem, an efficient two-step iterative algorithm is developed. Experiments on several UCI datasets and face image datasets validate the effectiveness and efficiency of our MKC algorithm.

Active learning via query synthesis and nearest neighbour search

Abstract: Active learning has received great interests from researchers due to its ability to reduce the amount of supervision required for effective learning. As the core component of active learning algorithms, query synthesis and pool-based sampling are two main scenarios of querying considered in the literature. Query synthesis features low querying time, but only has limited applications as the synthesized query might be unrecognizable to human oracle. As a result, most efforts have focused on pool-based sampling in recent years, although it is much more time-consuming. In this paper, we propose new strategies for a novel querying framework that combines query synthesis and pool-based sampling. It overcomes the limitation of query synthesis, and has the advantage of fast querying. The basic idea is to synthesize an instance close to the decision boundary using labelled data, and then select the real instance closest to the synthesized one as a query. For this purpose, we propose a synthesis strategy, which can synthesize instances close to the decision boundary and spreading along the decision boundary. Since the synthesis only depends on the relatively small labelled set, instead of evaluating the entire unlabelled set as many other active learning algorithms do, our method has the advantage of efficiency. In order to handle more complicated data and make our framework compatible with powerful kernel-based learners, we also extend our method to kernel version. Experiments on several real-world data sets show that our method has significant advantage on time complexity and similar performance compared to pool-based uncertainty sampling methods.

Feature and Region Selection for Visual Learning

Abstract: Visual learning problems such as object classification and action recognition are typically approached using extensions of the popular bag-of-words (BoW) model. Despite its great success, it is unclear what visual features the BoW model is learning: Which regions in the image or video are used to discriminate among classes? Which are the most discriminative visual words? Answering these questions is fundamental for understanding existing BoW models and inspiring better models for visual recognition. To answer these questions, this paper presents a method for feature selection and region selection in the visual BoW model. This allows for an intermediate visualization of the features and regions that are important for visual learning. The main idea is to assign latent weights to the features or regions, and jointly optimize these latent variables with the parameters of a classifier (e.g., support vector machine). There are four main benefits of our approach: (1) Our approach accommodates non-linear additive kernels such as the popular $\chi^2$ and intersection kernel; (2) our approach is able to handle both regions in images and spatio-temporal regions in videos in a unified way; (3) the feature selection problem is convex, and both problems can be solved using a scalable reduced gradient method; (4) we point out strong connections with multiple kernel learning and multiple instance learning approaches. Experimental results in the PASCAL VOC 2007, MSR Action Dataset II and YouTube illustrate the benefits of our approach.

Instance Annotation via Optimal BoW for Weakly Supervised Object Localization

Abstract: In this paper, we aim at irregular-shape object localization under weak supervision. With over-segmentation, this task can be transformed into multiple-instance context. However, most multiple-instance learning methods only emphasize single most positive instance in a positive bag to optimize bag-level classification, and leads to imprecise or incomplete localization. To address this issue, we propose a scheme for instance annotation, where all of the positive instances are detected by labeling each instance in each positive bag. Inspired by the successful application of bag-of-words (BoW) to feature representation, we leverage it at instance-level to model the distributions of the positive class and negative class, and then incorporate the BoW learning and instance labeling in a single optimization formulation. We also demonstrate that the scheme is well suited to weakly supervised object localization of irregular-shape. Experimental results validate the effectiveness both for the problem of generic instance annotation and for the application of weakly supervised object localization compared to some existing methods.

Feature and Region Selection for Visual Learning

Abstract: Visual learning problems, such as object classification and action recognition, are typically approached using extensions of the popular bag-of-words (BoWs) model. Despite its great success, it is unclear what visual features the BoW model is learning. Which regions in the image or video are used to discriminate among classes? Which are the most discriminative visual words? Answering these questions is fundamental for understanding existing BoW models and inspiring better models for visual recognition. To answer these questions, this paper presents a method for feature selection and region selection in the visual BoW model. This allows for an intermediate visualization of the features and regions that are important for visual learning. The main idea is to assign latent weights to the features or regions, and jointly optimize these latent variables with the parameters of a classifier (e.g., support vector machine). There are four main benefits of our approach: 1) our approach accommodates non-linear additive kernels, such as the popular $\chi ^{2}$ and intersection kernel; 2) our approach is able to handle both regions in images and spatio-temporal regions in videos in a unified way; 3) the feature selection problem is convex, and both problems can be solved using a scalable reduced gradient method; and 4) we point out strong connections with multiple kernel learning and multiple instance learning approaches. Experimental results in the PASCAL VOC 2007, MSR Action Dataset II and YouTube illustrate the benefits of our approach.

GMC: Graph-Based Multi-View Clustering

Abstract: Multi-view graph-based clustering aims to provide clustering solutions to multi-view data. However, most existing methods do not give sufficient consideration to weights of different views and require an additional clustering step to produce the final clusters. They also usually optimize their objectives based on fixed graph similarity matrices of all views. In this paper, we propose a general G raph-based M ulti-view C lustering (GMC) to tackle these problems. GMC takes the data graph matrices of all views and fuses them to generate a unified graph matrix. The unified graph matrix in turn improves the data graph matrix of each view, and also gives the final clusters directly. The key novelty of GMC is its learning method, which can help the learning of each view graph matrix and the learning of the unified graph matrix in a mutual reinforcement manner. A novel multi-view fusion technique can automatically weight each data graph matrix to derive the unified graph matrix. A rank constraint without introducing a tuning parameter is also imposed on the graph Laplacian matrix of the unified matrix, which helps partition the data points naturally into the required number of clusters. An alternating iterative optimization algorithm is presented to optimize the objective function. Experimental results using both toy data and real-world data demonstrate that the proposed method outperforms state-of-the-art baselines markedly.

SG-One: Similarity Guidance Network for One-Shot Semantic Segmentation

Abstract: One-shot image semantic segmentation poses a challenging task of recognizing the object regions from unseen categories with only one annotated example as supervision. In this article, we propose a simple yet effective similarity guidance network to tackle the one-shot (SG-One) segmentation problem. We aim at predicting the segmentation mask of a query image with the reference to one densely labeled support image of the same category. To obtain the robust representative feature of the support image, we first adopt a masked average pooling strategy for producing the guidance features by only taking the pixels belonging to the support image into account. We then leverage the cosine similarity to build the relationship between the guidance features and features of pixels from the query image. In this way, the possibilities embedded in the produced similarity maps can be adopted to guide the process of segmenting objects. Furthermore, our SG-One is a unified framework that can efficiently process both support and query images within one network and be learned in an end-to-end manner. We conduct extensive experiments on Pascal VOC 2012. In particular, our SG-One achieves the mIoU score of 46.3%, surpassing the baseline methods.

Multi-view Clustering: A Survey

Abstract: In the big data era, the data are generated from different sources or observed from different views. These data are referred to as multi-view data. Unleashing the power of knowledge in multi-view data is very important in big data mining and analysis. This calls for advanced techniques that consider the diversity of different views, while fusing these data. Multi-view Clustering (MvC) has attracted increasing attention in recent years by aiming to exploit complementary and consensus information across multiple views. This paper summarizes a large number of multi-view clustering algorithms, provides a taxonomy according to the mechanisms and principles involved, and classifies these algorithms into five categories, namely, co-training style algorithms, multi-kernel learning, multiview graph clustering, multi-view subspace clustering, and multi-task multi-view clustering. Therein, multi-view graph clustering is further categorized as graph-based, network-based, and spectral-based methods. Multi-view subspace clustering is further divided into subspace learning-based, and non-negative matrix factorization-based methods. This paper does not only introduce the mechanisms for each category of methods, but also gives a few examples for how these techniques are used. In addition, it lists some publically available multi-view datasets. Overall, this paper serves as an introductory text and survey for multi-view clustering.

A study of graph-based system for multi-view clustering

Abstract: This paper studies clustering of multi-view data, known as multi-view clustering. Among existing multi-view clustering methods, one representative category of methods is the graph-based approach. Despite its elegant and simple formulation, the graph-based approach has not been studied in terms of (a) the generalization of the approach or (b) the impact of different graph metrics on the clustering results. This paper extends this important approach by first proposing a general Graph-Based System (GBS) for multi-view clustering, and then discussing and evaluating the impact of different graph metrics on the multi-view clustering performance within the proposed framework. GBS works by extracting data feature matrix of each view, constructing graph matrices of all views, and fusing the constructed graph matrices to generate a unified graph matrix, which gives the final clusters. A novel multi-view clustering method that works in the GBS framework is also proposed, which can (1) construct data graph matrices effectively, (2) weight each graph matrix automatically, and (3) produce clustering results directly. Experimental results on benchmark datasets show that the proposed method outperforms state-of-the-art baselines significantly.