Qingcheng Zeng

The Steiner [ ]-eccentricity of a vertex [ ]-set [ ] is the maximum Steiner [ ]-distance among all vertex sets containing [ ]. In this paper, we introduce the Vicsek fractal [ ] and its corresponding network sequence. Using the finite pattern method, we then deal with the average Steiner eccentricities of Vicsek fractal and give the exact value of them. Furthermore, the asymptotic formula of the average Steiner eccentricities of Vicsek networks is derived.

Substitution networks can be constructed inductively based on multiple colored initial graphs. In this paper, we obtain a relationship between the row sums of power of a nonnegative matrix and its spectral radius. Furthermore, we investigate the substitution networks with colored initial directed trees and obtain their Hausdorff dimensions introduced by Xi et al.

In 2023, Xi et al. introduced the Hausdorff dimension of a family of networks which inspired by the potential theoretic methods in fractal geometry. In this paper, we will construct a class of colored substitution networks and obtain its Hausdorff dimension using the self-similarity.

Many complex networks have scale-free and small-world effects. In this paper, a family of evolving networks is constructed modeled by a non-symmetric self-similar planar fractal, using the encoding method in fractal geometry. Based on the self-similar structure, we study the degree distribution, clustering coefficient and average path length of our evolving network to verify their scale-free and small-world characteristics.

In this paper, we investigate the equivalence of connectedness for the Sierpinski-like sponge and skeleton networks, and find out the relation between the geodesic distance on the sponge and renormalized shortest path distance on the skeleton networks. Furthermore, under some assumption on the IFS, we obtain the comparability of the Manhattan distance and the geodesic distance on the sponge.

We obtain the Hausdorff dimension of the skeleton networks of Sierpiński carpet by the self-similarity and induction, which will not touch networks.

Fine tuning self supervised pretrained models using pseudo labels can effectively improve speech recognition performance. But, low quality pseudo labels can misguide decision boundaries and degrade performance. We propose a simple yet effective strategy to filter low quality pseudo labels to alleviate this problem. Specifically, pseudo-labels are produced over the entire training set and filtered via average probability scores calculated from the model output. Subsequently, an optimal percentage of utterances with high probability scores are considered reliable training data with trustworthy labels. The model is iteratively updated to correct the unreliable pseudo labels to minimize the effect of noisy labels. The process above is repeated until unreliable pseudo abels have been adequately corrected. Extensive experiments on LibriSpeech show that these filtered samples enable the refined model to yield more correct predictions, leading to better ASR performances under various experimental settings.

The past ten years have witnessed the rapid development of text-based intent detection, whose benchmark performances have already been taken to a remarkable level by deep learning techniques. However, automatic speech recognition (ASR) errors are inevitable in real-world applications due to the environment noise, unique speech patterns and etc, leading to sharp performance drop in state-of-the-art text-based intent detection models. Essentially, this phenomenon is caused by the semantic drift brought by ASR errors and most existing works tend to focus on designing new model structures to reduce its impact, which is at the expense of versatility and flexibility. Different from previous one-piece model, in this paper, we propose a novel and agile framework called CR-ID for ASR error robust intent detection with two plug-and-play modules, namely semantic drift calibration module (SDCM) and phonemic refinement module (PRM), which are both model-agnostic and thus could be easily integrated to any existing intent detection models without modifying their structures. Experimental results on SNIPS dataset show that, our proposed CR-ID framework achieves competitive performance and outperform all the baseline methods on ASR outputs, which verifies that CR-ID can effectively alleviate the semantic drift caused by ASR errors.