Publications Feed

Our paper “Large-Dimensional Characterization of Robust Linear Discriminant Analysis” with Nicolas Auguin and Matthew McKay (HKUST) has just appeared in IEEE Transactions on Signal Processing.

The work focuses on a robust version of linear discriminant analysis (LDA) classifiers, extensively used in machine learning. To account for potential spurious (e.g., non-Gaussian) or mislabeled observations in the training data (i.e., outlying samples), we here propose and characterise the large-dimensional performance (when both the numbers of samples and variables are large) of LDA classifiers employing robust regularised M-estimators of the covariance matrix, essential to the design of a robust discrimination rule when the dataset is corrupted with outliers.

Our paper “The road to 6G: Ten physical layer challenges for communications engineers” with M. Matthaiou, O. Yurduseven, H. Q. Ngo, S. L. Cotton and V. F. Fusco (QUB) has just appeared in IEEE Communications Magazine.

This magazine article discusses some of the most relevant open problems and challenges in the road ahead towards the realisation of 6G wireless networks. It stems from a collaborative effort between members (including myself) of the Center for Wireless Innovation (CWI) at Queen’s University Belfast.

Our paper “RocaSec: A standalone GUI-based package for robust co-evolutionary analysis of proteins” with Ahmed A. Quadeer and Matthew R. McKay (HKUST) has been accepted for publication in Bioinformatics.

This work presents a standalone software package with a convenient graphical user interface for co-evolutionary analyses of proteins. The software, along with several data examples for HIV and HCV proteins is now publicly accessible at https://github.com/ahmedaq/RocaSec

Our paper “Extreme eigenvalue distributions of Jacobi ensembles: New exact representations, asymptotics and finite size corrections”, with Laureano Moreno-Pozas and Matthew McKay (HKUST) just appeared in Nuclear Physics B.

This work presents theoretical advances in the field of high-dimensional random matrix theory and, in particular, of relevance to the applied statistics community. Statistical distributions for the extreme eigenvalues of Jacobi random matrices are derived and, for the first time, we bridge the gap between exact (finite-dimensional) results and asymptotic results, providing finite-size corrections to the asymptotic results. These fundamental results find application in many cross-disciplinary problems including statistical hypothesis testing, multi-antenna wireless communication systems with co-channel interference and quantum conductance in mesoscopic physics.

Our paper “Asymptotics of eigenstructure of sample correlation matrices for high-dimensional spiked models” with Iain Johnstone and Jeha Yang (Stanford) and Matthew McKay (HKUST) has been accepted for publication in Statistica Sinica.

This paper has been selected to be part of Sinica’s invited special session at JSM 2020.  Sinica hold an invited session at JSM (a huge stats meeting) only once every 2 years. Last time there were only 3 papers in the Sinica special session.  

The work characterises the statistics of sample correlation matrices by providing the asymptotic limits and fluctuations (central limit theorems) for the eigenvalues and eigenvectors of high-dimensional correlation matrices. These are fundamental to a myriad of problems based on principal component analysis (PCA), paving the way towards improved PCA in high dimensions, with relevant ramifications into data science and machine learning, particularly in inference problems from large-dimensional data.

Our paper “New approximation to the distribution of positive RVs applied to Gaussian quadratic forms” with Pablo (visiting student from Malaga) and his PhD supervisors has just appeared in IEEE Signal Processing Letters.   Well done Pablo!!

The paper proposes a new approach to the classical problem of approximating the distribution of a random variable, and this is applied to Gaussian quadratic forms. Compared to other methods, the new approximation always yield a valid approximating distribution and it is computationally more stable. The potential applications are numerous in engineering, wireless communications, signal processing and applied statistics; some examples are the characterisation of hypothesis testing, signal detection, and performance analysis of wireless channels.

Our paper “Sub-dominant principal components inform new vaccine targets for HIV Gag” with Faraz, Ahmed and Matthew (HKUST) has just appeared in Bioinformatics.

The paper proposes a computational method to accurately infer networks of interacting sites in viral proteins such as HIV Gag from patient-derived data sequences. We reveal that certain networks that appear important for vaccine design purposes are not accurately reflected by previous methods, based only on the dominant PCs. Rather, these networks are encoded jointly by both dominant and sub-dominant PCs. The new method is able to identify a network of interacting sites of HIV Gag that associated very strongly with viral control. Based on this, several new candidates for a potent T-cell-based HIV vaccine are put forward.

Our paper “Co-evolution networks of HIV/HCV are modular with direct association to structure and function”, with A. A. Quadeer and M. R. McKay, has been accepted for publication in PLOS Computational Biology. This is a multi-disciplinary research effort which combines new computational methods based on random matrix theory (RMT) with biology and viral evolutionary concepts to establish exciting insights into the mutational co-evolutionary networks employed by both HIV and HCV to evade immune pressure while remaining functionally fit. The results have important implications as they could enlighten the design of new vaccines against these viruses.  A preprint of this work is available on the link above.

Our paper “Large-dimensional behavior of regularized Maronna’s M-estimators of covariance matrices“, with N. Auguin, M. R. McKay and R. Couillet, has just been accepted for publication in IEEE Transactions on Signal Processing.

Our works “Largest Eigenvalue Distribution of Noncircularly-symmetric Wishart-type Matrices with Application to Hoyt-faded MIMO Communications”, and “Exact Statistical Characterization of 2×2 Gram Matrices with Arbitrary Variance Profile”, have recently been accepted for publication in the IEEE Transactions on Vehicular Technology.