We analyze two different estimation algorithms for dynamic average consensus in sensing and communication networks, a proportional algorithm proportional-integral algorithm. investigate the stability properties of these estimators under changing inputs network topologies as well their convergence constant or slowly-varying inputs. In doing so, we discover that more complex has performance benefits over simpler

Abstract Motivation: In silico methods provide efficient ways to predict possible interactions between drugs and targets. Supervised learning approach, bipartite local model (BLM), has recently been shown be effective in prediction of drug–target interactions. However, for drug-candidate compounds or target-candidate proteins that currently have no known available, its pure ‘local’ is not able learned hence BLM may fail make correct when involving such kind new candidates. Results: We...

Cooperating mobile sensors can be used to model environmental functions such as the temperature or salinity of a region ocean. In this paper, we adopt an optimal filtering approach fusing local sensor data into global environment. Our is based on use proportional-integral (PI) average consensus estimators, whereby information from each diffuses through communication network. As result, scalable and fully decentralized, allows changing network topologies anonymous agents added subtracted at...

Abstract Background: Identifying disease genes from human genome is an important but challenging task in biomedical research. Machine learning methods can be applied to discover new based on the known ones. Existing machine typically use as positive training set P and unknown negative N (non-disease gene does not exist) build classifiers identify genes. However, such kind of actually built a noisy there itself. As result, do perform well they could be. Result: Instead treating examples N, we...

An increasing number of genes have been experimentally confirmed in recent years as causative to various human diseases. The newly available knowledge can be exploited by machine learning methods discover additional unknown that are likely associated with In particular, positive unlabeled (PU learning) methods, which require only a training set P (confirmed disease genes) and an U (the candidate instead negative N, shown effective uncovering new the current scenario. Using single source data...

Phenotypically similar diseases have been found to be caused by functionally related genes, suggesting a modular organization of the genetic landscape human that mirrors modularity observed in biological interaction networks. Protein complexes, as molecular machines integrate multiple gene products perform functions, express underlying protein-protein As such, protein complexes can useful for interrogating networks phenome and interactome elucidate gene-phenotype associations diseases.We...

We describe distributed estimation algorithms that allow robots in a communication network to maintain estimates of summary statistics describing the shape swarm. show these estimators, combined with motion controllers implemented on each robot, result swarm formation being driven desired values presence changing topology and addition deletion

ABSTRACT Motivation Proper prioritization of candidate genes is essential to the genome-based diagnostics a range genetic diseases. However, it highly challenging task involving limited and noisy knowledge genes, diseases their associations. While number computational methods have been developed for disease gene task, performance largely by manually crafted features, network topology, or pre-defined rules data fusion. Results Here, we propose novel graph convolutional network-based method,...

Deep neural networks (DNNs) have become state-of-the-art in many application domains. The increasing complexity and cost for building these models demand means protecting their intellectual property (IP). This paper presents a novel DNN framework that optimizes the robustness of embedded watermarks. Our method is originated from fault attacks. Different prior end-to-end watermarking approaches, we only modify tiny subset weights to embed watermark, which also facilities better control model...

Multi-objective Bayesian optimization (MOBO) has shown promising performance on various expensive multi-objective problems (EMOPs). However, effectively modeling complex distributions of the Pareto optimal solutions is difficult with limited function evaluations. Existing set learning algorithms may exhibit considerable instability in such scenarios, leading to significant deviations between obtained solution and (PS). In this paper, we propose a novel Composite Diffusion Model based Set...

Outlier detection has many important applications in the field of fraud detection, network robustness analysis and intrusion detection. Some researches have utilized neural to solve problem because it advantage powerful modeling ability. In this paper, we propose a RBF model using subtractive clustering algorithm for selecting hidden node centers, which can achieve faster training speed. meantime, was trained with regularization term so as minimize variances nodes layer perform more accurate...

We investigate the use of intrinsic spectral analysis (ISA) for query-by-example spoken term detection (QbE-STD). In task, queries and test utterances in an audio archive are converted to ISA features, dynamic time warping is applied match feature sequence each query with those utterances. Motivated by manifold learning, has been proposed recover from untranscribed a set nonlinear basis functions speech manifold, shown improved phonetic separability inherent speaker independence. Due...

Cross-speaker style transfer in speech synthesis aims at transferring a from source speaker to synthesized of target speaker's timbre. In most previous methods, the fine-grained prosody features often represent average style, similar one-to-many problem(i.e., multiple variations correspond same text). response this problem, strength-controlled semi-supervised extractor is proposed disentangle content and timbre, improving representation interpretability global embedding, which can alleviate...

An outlier is the object which very different from rest of dataset on some measure. Finding such exception has received much attention in data mining field. In this paper, we propose a KNN based detection algorithm consisted two phases. Firstly, it partitions into several clusters and then each cluster, calculates Kth nearest neighborhood for to find outliers. addition, pruning scheme used our algorithm. It can effectively avoid frequent passing entire unnecessary computations. Experimental...

In this paper, we propose a partial sequence matching based symbolic search (SS) method for the task of language independent query-by-example spoken term detection. One main drawback conventional SS approach is high miss rate long queries. This due to variations in symbol representation query and audios, especially scenario. The successful with its instances audio becomes exponentially more difficult as grows longer. To reduce rate, strategy, which all phone sequences are used instances....

Traditional online learning algorithms are designed for vector data only, which assume that the labels of all training examples provided. In this paper, we study graph classification where only limited nodes chosen labelling by selective sampling. Particularly, first adapt a spectral-based regularization technique to derive novel linear algorithm can handle data, although it still queries and thus is not preferred, as typically time-consuming. To address issue, then propose new...

Although few-shot meta learning has been extensively studied in machine community, the fast adaptation towards new tasks remains a challenge scenario. The neuroscience research reveals that capability of evolving neural network formulation is essential for task adaptation, which broadly recent meta-learning researches. In this paper, we present novel forward-backward framework (FBM) to facilitate model generalization from perspective, i.e., neuron calibration. particular, FBM models neurons...

Through the development of powerful algorithms and design tools, deep neural networks (DNNs) have recently approached or even surpassed human-level performance in many real-world applications. Nowadays, since a product-level DNN modeling requires large amount training data expensive computing resources thus models are considered as valuable data, protecting intellectual property (IP) builders becomes an important problem security domain. In this paper, we propose novel watermarking approach...

Multi-Task Learning (MTL) can enhance a classifier's generalization performance by learning multiple related tasks simultaneously. Conventional MTL works under the offline or batch setting, and suffers from expensive training cost poor scalability. To address such inefficiency issues, online techniques have been applied to solve problems. However, most existing algorithms of constrain task relatedness into presumed structure via single weight matrix, which is strict restriction that does not...

Traditional online learning for graph node classification adapts regularization into ridge regression, which may not be suitable when data is adversarially generated. To solve this issue, we propose a more general min-max optimization framework classification. The derived algorithm can achieve regret compared with the optimal linear model found offline. However, assumes that label provided every node, while scare and labeling usually either too time-consuming or expensive in real-world...

Herein, to address the challenges faced by Automatic Guided Vehicles (AGVs) in construction site environments, including heavy vehicle loads, extensive road search areas, and randomly distributed obstacles, this paper presents a hierarchical trajectory planning algorithm that combines coarse precise planning. In first-level planning, lateral longitudinal sampling is performed based on environment constraints. A multi-criteria cost function designed, taking into account factors such as...

This paper studies the effect of communication link weighting schemes and long-range connections in speeding convergence to consensus on average inputs agents a sensor network. Linear programming LMI solutions are provided for problem centralized optimization weights considering possibility intermittent bounded delays. Heuristic that can be implemented distributed fashion also studied. Consensus time may reduced by introducing interactions, creating "small world" network