Neural sequence-to-sequence models for dialog systems suffer from the problem of favoring uninformative and non replier-specific responses due to lacking of the global and relevant information guidance. The existing methods model the generation process by leveraging the neural variational network with simple Gaussian. However, the sampled information from latent space usually becomes useless due to the KL divergence vanishing issue, and the highly abstractive global variables easily dilute the personal features of replier, leading to a non replier-specific response. Therefore, a novel Semi-Supervised Stable Variational Network (SSVN) is proposed to address these issues. We use a unit hypersperical distribution, namely the von Mises-Fisher (vMF), as the latent space of a semi-supervised model, which can obtain the stable KL performance by setting a fixed variance and hence enhance the global information representation. Meanwhile, an unsupervised extractor is introduced to automatically distill the replier-tailored feature which is then injected into a supervised generator to encourage the replier-consistency. Experimental results on two large conversation datasets show that our model outperforms the competitive baseline models significantly, and can generate diverse and replier-specific responses.
When designing the horn of the ultrasonic transducer for rotary machining, the traditional analytical method cannot consider the complex assembly cross-section of the horn, resulting in a large final simulation error and design error. At the same time, when the design diameter of the transducer is too large, the size of the back mass obtained by the traditional analytical method is too thin. This paper combines the four terminal network method and the finite element method, selects the sandwich piezoelectric transducer, takes the ER spring chuck commonly used in rotary machining as an example, derives the transmission matrix corresponding to the geometric cross-section shape, and then uses the four-terminal network method and the traditional analytical method to design two sets of stepped ultrasonic transducers with the design frequency of 22 kHz for analysis and comparison. Finally, through the analysis, it is verified that the ultrasonic transducer designed by the four-terminal network method has a performance closer to the theoretical design.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.