Graph kernels combined with the neural network on protein classification

At present, most of the researches on protein classification are based on graph kernels. The essence of graph kernels is to extract the substructure and use the similarity of substructures as the kernel values. In this paper, we propose a novel graph kernel named vertex-edge similarity kernel (VES kernel) based on mixed matrix, the innovation point of which is taking the adjacency matrix of the graph as the sample vector of each vertex and calculating kernel values by finding the most similar vertex pair of two graphs. In addition, we combine the novel kernel with the neural network and the experimental results show that the combination is better than the existing advanced methods.

A New Recurrence-Network-Based Time Series Analysis Approach for Characterizing System Dynamics

In this paper, a novel analysis method based on recurrence networks is proposed to characterize the evolution of dynamical systems. Through phase space reconstruction, a time series was transformed into a high-dimensional recurrence network and a corresponding low-dimensional recurrence network, respectively. Then, two appropriate statistics, the correlation coefficient of node degrees (CCND) and the edge similarity, were proposed to unravel the evolution properties of the considered signal. Through the investigation of the time series with distinct dynamics, different patterns in the decline rate of the CCND at different network dimensions were observed. Interestingly, an exponential scaling emerged in the CCND analysis for the chaotic time series. Moreover, it was demonstrated that the edge similarity can further characterize dynamical systems and provide detailed information on the studied time series. A method based on the fluctuation of edge similarities for neighboring edge groups was proposed to determine the number of groups that the edges should be partitioned into. Through the analysis of chaotic series corrupted by noise, it was demonstrated that both the CCND and edge similarity derived from different time series are robust under additive noise. Finally, the application of the proposed method to ventricular time series showed its effectiveness in differentiating healthy subjects from ventricular tachycardia (VT) patients.

Evaluation of Object Surface Edge Profiles Detected with a 2-D Laser Scanning Sensor

Canopy edge profile detection is a critical component of plant recognition in variable-rate spray control systems. The accuracy of a high-speed 270° radial laser sensor was evaluated in detecting the surface edge profiles of six complex-shaped objects. These objects were toy balls with a pink smooth surface, light brown rectangular cardboard boxes, black and red texture surfaced basketballs, white smooth cylinders, and two different sized artificial plants. Evaluations included reconstructed three-dimensional (3-D) images for the object surfaces with the data acquired from the laser sensor at four different detection heights (0.25, 0.50, 0.75, and 1.00 m) above each object, five sensor travel speeds (1.6, 2.4, 3.2, 4.0, and 4.8 km h−1), and 8 to 15 horizontal distances to the sensor ranging from 0 to 3.5 m. Edge profiles of the six objects detected with the laser sensor were compared with images taken with a digital camera. The edge similarity score (ESS) was significantly affected by the horizontal distances of the objects, and the influence became weaker when the objects were placed closer to each other. The detection heights and travel speeds also influenced the ESS slightly. The overall average ESS ranged from 0.38 to 0.95 for all the objects under all the test conditions, thereby providing baseline information for the integration of the laser sensor into future development of greenhouse variable-rate spray systems to improve pesticide, irrigation, and nutrition application efficiencies through watering booms.

Optimizing Ontology Alignments by Using Genetic Algorithms

In this work we present GOAL (Genetics for Ontology Align- ments) a new approach to compute the optimal ontology alignment func- tion for a given ontology input set. Although this problem could be solved by an exhaustive search when the number of similarity measures is low, our method,is expected to scale better for a high number,of measures. Our approach is a genetic algorithm which is able to work with several goals: maximizing the alignment precision, maximizing the alignment re- call, maximizing the f-measure or reducing the number of false positives. Moreover, we test it here by combining some cutting-edge similarity mea- sures over a standard benchmark, and the results obtained show several advantages in relation to other techniques. Key words: ontology alignment; genetic algorithms; semantic integra-