Search

Can the robot think? This is a major topic of artificial intelligence and has always attracted the attention and exploration of many researchers. Since Searle proposed the Chinese House Thought Experiment in 1980, researchers have conducted extensive research on whether or not a formal symbolic system itself can understand the meaning of symbols, and have obtained many insights. From the perspective of cognitive science, this paper studies how robots understand knowledge, and proposes a general model that expresses robots’ understanding of common concepts (the simplest form of knowledge) to describe the process by which robots understand some simple knowledge. We actually find a new way for robots to understand complex knowledge. In this paper, we also illustrate our method by an example of robot’s understanding the concept of “pen”.

Knowledge representation and reasoning is one of the most important topics in the field of artificial intelligence. Classical 2-valued logic is often used for representation and reasoning of precise knowledge. However, in practical applications, intelligent systems need to understand and handle knowledge involving time and fuzzy concepts, which classical 2-valued logic cannot do well. To this end, in this paper we propose a 2n-valued temporal logic for representing and reasoning of temporal, fuzzy knowledge. Moreover, we discuss the characteristics of the logic, including some of theorems that hold in the classical logic as well as some unique to the logical proposed. Finally, we give an example of the impact of seasonal cooling on influenza to show how our logic can be used in practice.

It is crucial to appropriately deploy Roadside Units (RSUs) to guarantee the QoS of Vehicle-to-Infrastructure communication. Sarubbi et al. proposed the Delta-r algorithm by using relative contact time to meet the \(\triangle _{\rho _{2}}^{\rho _{1}}\)-Deployment, which is a metric for evaluating the performance of VANET. However, the “false high” relative contact time might play a negative effect on decision-making. In this paper, an improved algorithm Delta-uc, which is based on Useful Contribution, is presented. It avoids the negative effect of “extra” contact time by retaining only the useful relative trip duration of a vehicle at an urban cell. In addition, based on the Delta-uc algorithm, an effective recombination operator is designed, and a parthenogenetic algorithm UCPGA is proposed to solve the deployment problem. Compared with algorithms Delta-r and Delta-GA, in many \(\triangle _{\rho _{2}}^{\rho _{1}}\)-Deployments, the Delta-uc and UCPGA algorithms respectively required fewer RSUs, which were proved by the experiments on the realistic mobility trace of Cologne, Germany.

In this paper, the 2-Species Duplication-Loss Small Phylogeny Problem is studied. By introducing an alignment algorithm, a labeling algorithm and three smart mutate operators, a genetic algorithm G2SP is presented. Algorithm G2SP adopts the method of combining the general operator and the smart ones. The general operator maintains the population diversity effectively, while the smart ones improve population convergence and make it evolve to the optimal solution more quickly. The tRNA and rRNA gene data of six kinds of real bacterium were used to test the performance of algorithms. The experimental results indicate that the G2SP algorithm can get fewer evolution cost than the PBLB algorithm proposed by Holloway et al., and it is an effective method for solving the 2-species duplication-loss small phylogeny problem.

The minimum error correction (MEC) model is one of the important computational models for determining haplotype information from sequencing data, i.e., single individual single nucleotide polymorphism (SNP) haplotyping, haplotype reconstruction or haplotype assembly. Due to the NP-hardness of the model, a fast and accurate enumeration algorithm is proposed for solving it. The presented algorithm reconstructs the SNP sites of a pair of haplotypes one after another. It enumerates two kinds of SNP values, i.e., (0 1)T and (1 0)T, for the SNP site being reconstructed, and chooses the one with more support coming from the SNP fragments that are covering the corresponding SNP site. The experimental comparisons were conducted among the presented algorithm, the FAHR, the Fast Hare and the DGS algorithms. The results prove that our algorithm can get higher reconstruction rate than the other three algorithms.

Resolution principle is an important rule of inference in theorem proving. Model counting using extension rule is considered as a counterpart to resolution-based methods for model counting. Based on the exact method, this paper proposes two approximate model counting algorithms, and proves the time complexity of the algorithms. Experimental results show that they have good performance in the accuracy and efficiency.