Transfer Neural Trees: Semi-Supervised Heterogeneous Domain Adaptation and Beyond.

Heterogeneous domain adaptation (HDA) addresses the task of associating data not only across dissimilar domains but also described by different types of features. Inspired by the recent advances of neural networks and deep learning, we propose a deep leaning model of Transfer Neural Trees (TNT), which jointly solves cross-domain feature mapping, adaptation, and classification in a unified architecture. As the prediction layer in TNT, we introduce Transfer Neural Decision Forest (Transfer- NDF), which is able to learn the neurons in TNT for adaptation by stochastic pruning. In order to handle semi-supervised HDA, a unique embedding loss term is introduced to TNT for preserving prediction and structural consistency between labeled and unlabeled target-domain data. We further show that our TNT can be extended to zero shot learning for associating image and attribute data with promising performance. Finally, experiments on different classification tasks across features, datasets, and modalities would verify the effectiveness of our TNT.

Learning Sample Specific Weights for Late Fusion.

Late fusion is one of the most effective approaches to enhance recognition accuracy through combining prediction scores of multiple classifiers, each of which is trained by a specific feature or model. The existing methods generally use a fixed fusion weight for one classifier over all samples, and ignore the fact that each classifier may perform better or worse for different subsets of samples. In order to address this issue, we propose a novel sample specific late fusion (SSLF) method. Specifically, we cast late fusion into an information propagation process that diffuses the fusion weights of labeled samples to the individual unlabeled samples, and enforce positive samples to have higher fusion scores than negative samples. Upon this process, the optimal fusion weight for each sample is identified, while positive samples are pushed toward the top at the fusion score rank list to achieve better accuracy. In this paper, two SSLF methods are presented. The first method is ranking SSLF (R-SSLF), which is based on graph Laplacian with RankSVM style constraints. We formulate and solve the problem with a fast gradient projection algorithm; the second method is infinite push SSLF (I-SSLF), which combines graph Laplacian with infinite push constraints. I-SSLF is a l∞ norm constrained optimization problem and can be solved by an efficient alternating direction method of multipliers method. Extensive experiments on both large-scale image and video data sets demonstrate the effectiveness of our methods. In addition, in order to make our method scalable to support large data sets, the AnchorGraph model is employed to propagate information on a subset of samples (anchor points) and then reconstruct the entire graph to get the weights of all samples. To the best of our knowledge, this is the first method that supports learning of sample specific fusion weights for late fusion.

Efficient mining gapped sequential patterns for motifs in biological sequences.

BACKGROUND: Pattern mining for biological sequences is an important problem in bioinformatics and computational biology. Biological data mining yield impact in diverse biological fields, such as discovery of co-occurring biosequences, which is important for biological data analyses. The approaches of mining sequential patterns can discover all-length motifs of biological sequences. Nevertheless, traditional approaches of mining sequential patterns inefficiently mine DNA and protein data since the data have fewer letters and lengthy sequences. Furthermore, gap constraints are important in computational biology since they cope with irrelative regions, which are not conserved in evolution of biological sequences. RESULTS: We devise an approach to efficiently mine sequential patterns (motifs) with gap constraints in biological sequences. The approach is the Depth-First Spelling algorithm for mining sequential patterns of biological sequences with Gap constraints (termed DFSG). CONCLUSIONS: PrefixSpan is one of the most efficient methods in traditional approaches of mining sequential patterns, and it is the basis of GenPrefixSpan. GenPrefixSpan is an approach built on PrefixSpan with gap constraints, and therefore we compare DFSG with GenPrefixSpan. In the experimental results, DFSG mines biological sequences much faster than GenPrefixSpan.

Adaptive color feature extraction based on image color distributions.

This paper proposes an adaptive color feature extraction scheme by considering the color distribution of an image. Based on the binary quaternion-moment-preserving (BQMP) thresholding technique, the proposed extraction methods, fixed cardinality (FC) and variable cardinality (VC), are able to extract color features by preserving the color distribution of an image up to the third moment and to substantially reduce the distortion incurred in the extraction process. In addition to utilizing the earth mover's distance (EMD) as the distance measure of our color features, we also devise an efficient and effective distance measure, comparing histograms by clustering (CHIC). Moreover, the efficient implementation of our extraction methods is explored. With slight modification of the BQMP algorithm, our extraction methods are equipped with the capability of exploiting the concurrent property of hardware implementation. The experimental results show that our hardware implementation can achieve approximately a second order of magnitude improvement over the software implementation. It is noted that minimizing the distortion incurred in the extraction process can enhance the accuracy of the subsequent various image applications, and we evaluate the meaningfulness of the new extraction methods by the application to content-based image retrieval (CBIR). Our experimental results show that the proposed extraction methods can enhance the average retrieval precision rate by a factor of 25% over that of a traditional color feature extraction method.

A new approach to image copy detection based on extended feature sets.

Conventional image copy detection research concentrates on finding features that are robust enough to resist various kinds of image attacks. However, finding a globally effective fealure is difficult and, in many cases, domain dependent. Instead of imply extracting features from copyrighted images directly, we propose a new framework called the extended feature set for detecting copies of images. In our approach, virtual prior attacks are applied to copyrighted images to generate novel features, which serve as training data. The copy-detection problem can be solved by learning classifiers from the training data, thus, generated. Our approach can be integrated into existing copy detectors to further improve their performance. Experiment results demonstrate that the proposed approach can substantially enhance the accuracy of copy detection.