SearcharxivSearch

arXiv subjects

Aiden Williams

Publications and source records attributed to Aiden Williams.

2 recordsLinked to original sources

Generalized Nordhaus--Gaddum Inequalities for Eigenvalues

For a graph $G$, let $ \lambda_1(G)\ge \lambda_2(G)\ge \cdots \ge \lambda_n(G)$ denote the adjacency eigenvalues of $G$. We investigate the asymptotic maximum of \[ \lambda_i(G)+\lambda_j(\overline G) \] for fixed $i$ and $j$. We prove general bounds on $\lambda_i(G) + \lambda_{j}(\overline{G})$ for all pairs $(i, j)$ and also give general bounds on the related problem of minimizing $\lambda_{n-i+1}(G) + \lambda_{n-j+1}(\overline{G})$ for fixed $i$ and $j$. We prove that for all looped graphs $G$ on $n$ vertices, \[\lambda_1(G) + \lambda_2(\overline{G}) \le \frac87 n. \] Our method also gives a new short proof of the Nordhaus-Gaddum result for the spectral radius proved by Terpai that $\lambda_1(G) + \lambda_1(\overline{G}) \le \frac43n - 1$. We also show the close relation of these Nordhaus-Gaddum type problems to recent work on the maximum spectral gaps of graphs by Brooks, Linz and Lu.

math.CO

Analysis of Data Augmentation Methods for Low-Resource Maltese ASR

Recent years have seen an increased interest in the computational speech processing of Maltese, but resources remain sparse. In this paper, we consider data augmentation techniques for improving speech recognition for low-resource languages, focusing on Maltese as a test case. We consider three different types of data augmentation: unsupervised training, multilingual training and the use of synthesized speech as training data. The goal is to determine which of these techniques, or combination of them, is the most effective to improve speech recognition for languages where the starting point is a small corpus of approximately 7 hours of transcribed speech. Our results show that combining the data augmentation techniques studied here lead us to an absolute WER improvement of 15% without the use of a language model.

cs.CL