Search

Manifold learning has been widely used for dimensionality reduction and feature extraction of data recently. However, in the application of the related algorithms, it often suffers from noisy or unreliable data problems. For example, when the sample data have complex background, occlusions, and/or illuminations, the clustering of data is still a challenging task. To address these issues, we propose a family of novel algorithms for manifold regularized non-negative matrix factorization in this paper. In the algorithms, based on the alpha-beta-divergences, graph regularization with multiple segments is utilized to constrain the data transitivity in data decomposition. By adjusting two tuning parameters, we show that the proposed algorithms can significantly improve the robustness with respect to the images with complex background. The efficiency of the proposed algorithms is confirmed by the experiments on four different datasets. For different initializations and datasets, variations of cost functions and decomposition data elements in the learning are presented to show the convergent properties of the algorithms.

The D620N mutation in vacuolar protein sorting protein 35 (VPS35) gene has been identified to be linked to late onset familial Parkinson disease (PD). However, the pathophysiological roles of VPS35-D620N in PD remain unclear. Here, we generated the transgenic Caenorhabditis elegans overexpressing either human wild type or PD-linked mutant VPS35-D620N in neurons. C. elegans expressing VPS35-D620N, compared with non-transgenic controls, showed movement disorders and dopaminergic neuron loss. VPS35-D620N worms displayed more swimming induced paralysis but showed no defects in BSR assays, thus indicating the disruption of dopamine (DA) recycling back inside neurons. Moreover, VPS35 formed a protein interaction complex with DA transporter (DAT), RAB5, RAB11 and FAM21. In contrast, the VPS35-D620N mutant destabilized these interactions, thus disrupting DAT transport from early endosomes to recycling endosomes, and decreasing DAT at the cell surface. These effects together increased DA in synaptic clefts, and led to dopaminergic neuron degeneration and motor dysfunction. Treatment with reserpine significantly decreased the swimming induced paralysis in VPS35-D620N worms, as compared with vehicle treated VPS35-D620N worms. Our studies not only provide novel insights into the mechanisms of VPS35-D620N-induced dopaminergic neuron degeneration and motor dysfunction via disruption of DAT function and the DA signaling pathway but also indicate a potential strategy to treat VPS35-D620N-related PD and other disorders.

An attributed graph is a graph where nodes are associated with attributes describing their features. Clustering on an attributed graph is to detect clusters that have not only (1) cohesive structure; but also (2) homogeneous attribute values. We observe that in an attributed graph, different clusters tend to correlate to different attributes. For example, a cluster may show homogeneous values on attribute A but random values on attribute B, while for another cluster the situation could be the reverse. Existing attributed graph clustering methods ignore this phenomenon, thus leading to unsatisfactory clustering results. In this paper, we propose a novel attributed graph clustering framework called C o H omo to detect clusters with different attribute correlation patterns. The objective is to detect clusters that show both high structural cohesiveness and value homogeneity on some of the attributes. C o H omo defines and optimizes a correlation weight vector for each cluster to capture its correlation degrees to different node attributes. We formalize this problem as a bi-objective optimization problem and design an efficient heuristic optimizing method. The correlation weights for a cluster can be updated in two ways, synchronous updating or asynchronous updating, during the process of clustering. C o H omo is parameter-free and has near linear scalability. Theoretical analysis shows that all the results on the Pareto front generated by C o H omo are weak Pareto optimal solutions. Extensive experiments on attributed graphs based on several real world networks verify the effectiveness and show that our method outperforms the state-of-the-art method.

Cancer is a refractory disease, which brings physical, mental and economic sufferings on patients and their families. The activation of oncogenes and/or the deletion of tumor repressors result in uncontrolled tumor survival and proliferation. Therapeutic resistance and cancer recurrence are common issues during the course of clinical treatment. Since each single treatment can’t eradicate tumor cells completely, the development of combinatorial therapeutic regimen is in the spotlight for several decades. Combinatorial therapeutic regimen, a treatment modality that combines two or more therapeutic agents, targets key pathways in a characteristically synergistic or an additive manner. It aims to improve the efficacy and magnitude of therapeutic responses and reduce the likelihood of acquired resistance by using optimal drug dose with tolerable side effect. Currently, the combinatorial therapeutic regimen not only includes various chemicals, but also diverse therapeutic modalities which affected DNA damage repair, protein localization and degradation, post-translational modification, cell metabolism and apoptosis, as well as immune response.

Social influence analysis in microblogging services has attracted much attention in recent years. However, most previous studies were focused on measuring users’ (topical) influence. Little effort has been made to discern and quantify how a user is influenced. Specifically, the fact that user $i$ i retweets a tweet from author $j$ j could be either because $i$ i is influenced by $j$ j (i.e., $j$ j is a topical authority), or simply because he is “influenced” by the content (interested in the content). To mine such influence patterns, we propose a novel Bayesian factorization model, dubbed Influence Beta-Poisson Factorization (IBPF). IBPF jointly factorizes the retweet data and tweet content to quantify latent topical factors of user preference, author influence and content influence. It generates every retweet record according to the sum of two causing terms: one representing author influence, and the other one derived from content influence. To control the impact of the two terms, for each user IBPF generates a probability for each latent topic by Beta distribution, indicating how strongly the user cares about the topical authority of the author. We develop an efficient variational inference algorithm for IBPF. We demonstrate the efficacy of IBPF on two public microblogging datasets.

Imagine we are introducing a new product through a social network, where we know for each user in the network the function of purchase probability with respect to discount. Then, what discounts should we offer to those social network users so that, under a predefined budget, the adoption of the product is maximized in expectation? Although influence maximization has been extensively explored, this appealing practical problem still cannot be answered by the existing influence maximization methods. In this article, we tackle the problem systematically. We formulate the general continuous influence maximization problem, investigate the essential properties, and develop a general coordinate descent algorithmic framework as well as the engineering techniques for practical implementation. Our investigation does not assume any specific influence model and thus is general and principled. At the same time, using the most popularly adopted triggering model as a concrete example, we demonstrate that more efficient methods are feasible under specific influence models. Our extensive empirical study on four benchmark real-world networks with synthesized purchase probability curves clearly illustrates that continuous influence maximization can improve influence spread significantly with very moderate extra running time comparing to the classical influence maximization methods.

The Person Re-identification (ReID) task aims to match persons across cameras in a surveillance system. In the past few years, many researches are devoted to ReID and its performance has gained significant improvement. ReID models are usually trained as a joint framework comprising a person feature extractor and a classifier. However, there exists co-adaptation between the feature extractor and the classifier, which prevents the feature extractor from making effective and sufficient optimization and results in inferior retrieval performance. In this paper, we propose a very simple and effective training method, called DeAda, to decouple this co-adaptation. Our main motivation is to construct a series of weak classifiers during training by randomization of parameters, so that optimization on the feature extractor could be strengthened in the training stage. DeAda is easy, effective, and efficient, and could serve as a plug-and-play optimization tool for ReID models, without additional memory and time cost. We also analyze the theoretical property of DeAda and show that it could produce identical features for the same person under some simple assumptions. We demonstrate its effectiveness on three public ReID datasets: Market1501, DukeMTMC-reID and CUHK03 over different ReID models. With DeAda optimization, we finally obtain state-of-the-art results on all the three datasets.

Background: Parkinson’s disease (PD) is the second most common neurodegenerative disease and the most common movement disorder characterized by motor impairments resulting from midbrain dopamine neuron loss. Abnormalities in small pial arteries and arterioles, which are the primary pathways of local delivery of nutrients and oxygen in brain tissue, have been reported in many neurodegenerative diseases including PD. Mutations in LRRK2 cause genetic PD and contribute to sporadic PD. The most common PD-linked mutation LRRK2 G2019S contributes 20–47% of genetic forms of PD in Caucasian populations. The human LRRK2 G2019S transgenic mouse model displays PD-like movement impairment and was used to identify novel LRRK2 inhibitors, which provides a useful model for studying microvascular abnormalities in PD. Objectives: To investigate abnormalities in arteriolar cerebral blood volume (CBVa) in various brain regions using the inflow-based vascular-space occupancy (iVASO) MRI technique in LRRK2 mouse models of PD. Methods: Anatomical and iVASO MRI scans were performed in 5 female and 7 male nontransgenic (nTg), 3 female and 4 male wild-type (WT) LRRK2, and 5 female and 7 male G2019S-LRRK2 mice of 9 months of age. CBVa was calculated and compared in the substantia nigra (SN), olfactory cortex, and prefrontal cortex. Results: Compared to nTg mice, G2019S-LRRK2 mice showed decreased CBVa in the SN, but increased CBVa in the olfactory and prefrontal cortex in both male and female groups, whereas WT-LRRK2 mice showed no change in CBVa in the SN (male and female), the olfactory (female), and prefrontal (female) cortex, but a slight increase in CBVa in the olfactory and prefrontal cortex in the male group only. Conclusions: Alterations in the blood volume of small arteries and arterioles (CBVa) were detected in the G2019S-LRRK2 mouse model of PD. The opposite changes in CBVa in the SN and the cortex indicate that PD pathology may have differential effects in different brain regions. Our results suggest the potential value of CBVa as a marker for clinical PD studies.

User information behavior is an important factor affecting the promotion effect of enterprises’ innovative products on mobile social networks. To provide insights as to how enterprises can better promote innovative products, this paper conducts a quantitative study of its internal function. We introduce a network evolutionary game model based on the Bass model, first to describe the background of the mobile social network, then to simulate the diffusion process of the social network and the user’s decision-making game behavior during promotion of enterprises’ innovative products, and finally to simulate the promotion effect on innovative products and solve for the conditions for the best innovative product marketing effect and the model are simulated from multiple angles. Our simulation results show that mobile social network marketing in the initial stage of product promotion has a significant impact on the promotion effect of innovative products. When user feedback is poor, advertisement diffusion has an obvious effect on product promotion. In mobile social networks, the proportion of positive feedback from users has a greater impact on the product promotion effect, while the proportion of negative feedback has a smaller impact on the product promotion effect. The impact of negative reviews is time sensitive. According to the simulation results, we propose several suggestions to improve the promotion effect of enterprises’ innovative products on mobile social networks.