arXiv subjects

Sandipan Roy

Publications and source records attributed to Sandipan Roy.

9 recordsLinked to original sources

arXiv2025

A Doubled Adjacency Spectral Embedding Approach to Graph Clustering

Spectral clustering is a popular tool in network data analysis, with applications in a variety of scientific application areas. However, many studies have shown that classical spectral clustering does not perform well on certain network structures, particularly core-periphery networks. To improve clustering performance in core-periphery structures, Adjacency Spectral Embedding (ASE) has been introduced, which performs clustering via a network's adjacency matrix instead of the graph Laplacian. Despite its advantages in this setting, the optimal performance of ASE is limited to dense networks, whilst network data observed in practice is often sparse in nature. To address this limitation, we propose a new approach which we term Doubled Adjacency Spectral Embedding (DASE), motivated by the observation that the squared adjacency matrix will leverage the fewer connections in sparse structures more efficiently in clustering applications. Theoretical results establish that the resulting clustering algorithm enjoys good consistency properties when determining sparse community structure. The performance and general applicability of the proposed method is evaluated using extensive simulations on both directed and undirected networks. Our results highlight the improved clustering performance on both sparse and dense networks in the presence of core-periphery structures. We illustrate our proposed technique on real-world employment and transportation datasets.

stat.ME↗

arXiv2024

Large multi-response linear regression estimation based on low-rank pre-smoothing

Pre-smoothing is a technique aimed at increasing the signal-to-noise ratio in data to improve subsequent estimation and model selection in regression problems. However, pre-smoothing has thus far been limited to the univariate response regression setting. However, there are many scientific applications in which interest lies in multi-response regression problems, particularly when the number of responses is large. Motivated by this setting, this article proposes a technique for data pre-smoothing based on low-rank approximation. We establish theoretical results on the performance of the proposed methodology, which show that in this large-response setting, the proposed technique outperforms ordinary least squares estimation with the mean squared error criterion, whilst being computationally more efficient than alternative approaches such as reduced rank regression. We quantify our estimator's benefit empirically in a number of simulated experiments. We also demonstrate our proposed low-rank pre-smoothing technique on real data arising from the environmental and biological sciences.

stat.ME↗

arXiv2023

A stochastic network approach to clustering and visualising single-cell genomic count data

Important tasks in the study of genomic data include the identification of groups of similar cells (for example by clustering), and visualisation of data summaries (for example by dimensional reduction). In this paper, we develop a novel approach to these tasks in the context of single-cell genomic data. To do so, we propose to model the observed genomic data count matrix $\mathbf{X}\in\mathbb{Z}_{\geq0}^{p\times n}$, by representing these measurements as a bipartite network with multi-edges. Utilising this first-principles network model of the raw data, we cluster single cells in a suitably identified $d$-dimensional Laplacian Eigenspace (LE) via a Gaussian mixture model (GMM-LE), and employ UMAP to non-linearly project the LE to two dimensions for visualisation (UMAP-LE). This LE representation of the data-points estimates transformed latent positions (of genes and cells), under a latent position statistical model of nodes in a bipartite stochastic network. We demonstrate how transformations of these estimated latent positions can enable fine-grained clustering and visualisation of single-cell genomic data, by application to data from three recent genomics studies in different biological contexts. In each data application, clusters of cells independently learned by our proposed methodology are found to correspond to cells expressing specific marker genes that were independently defined by domain experts. In this validation setting, our proposed clustering methodology outperforms the industry-standard for these data. Furthermore, we validate components of the LE decomposition of the data by contrasting healthy cells from normal and at-risk groups in a machine-learning model, thereby identifying an LE cancer biomarker that significantly predicts long-term patient survival outcome in two independent validation cohorts with data from 1904 and 1091 individuals.

stat.ME↗

arXiv2019

Interpretable brain age prediction using linear latent variable models of functional connectivity

Neuroimaging-driven prediction of brain age, defined as the predicted biological age of a subject using only brain imaging data, is an exciting avenue of research. In this work we seek to build models of brain age based on functional connectivity while prioritizing model interpretability and understanding. This way, the models serve to both provide accurate estimates of brain age as well as allow us to investigate changes in functional connectivity which occur during the ageing process. The methods proposed in this work consist of a two-step procedure: first, linear latent variable models, such as PCA and its extensions, are employed to learn reproducible functional connectivity networks present across a cohort of subjects. The activity within each network is subsequently employed as a feature in a linear regression model to predict brain age. The proposed framework is employed on the data from the CamCAN repository and the inferred brain age models are further demonstrated to generalize using data from two open-access repositories: the Human Connectome Project and the ATR Wide-Age-Range.

stat.AP↗

arXiv2018

Likelihood Inference for Large Scale Stochastic Blockmodels with Covariates based on a Divide-and-Conquer Parallelizable Algorithm with Communication

We consider a stochastic blockmodel equipped with node covariate information, that is helpful in analyzing social network data. The key objective is to obtain maximum likelihood estimates of the model parameters. For this task, we devise a fast, scalable Monte Carlo EM type algorithm based on case-control approximation of the log-likelihood coupled with a subsampling approach. A key feature of the proposed algorithm is its parallelizability, by processing portions of the data on several cores, while leveraging communication of key statistics across the cores during each iteration of the algorithm. The performance of the algorithm is evaluated on synthetic data sets and compared with competing methods for blockmodel parameter estimation. We also illustrate the model on data from a Facebook derived social network enhanced with node covariate information.

stat.CO↗

arXiv2018

Hedging parameter selection for basis pursuit

In Compressed Sensing and high dimensional estimation, signal recovery often relies on sparsity assumptions and estimation is performed via $\ell_1$-penalized least-squares optimization, a.k.a. LASSO. The $\ell_1$ penalisation is usually controlled by a weight, also called "relaxation parameter", denoted by $λ$. It is commonly thought that the practical efficiency of the LASSO for prediction crucially relies on accurate selection of $λ$. In this short note, we propose to consider the hyper-parameter selection problem from a new perspective which combines the Hedge online learning method by Freund and Shapire, with the stochastic Frank-Wolfe method for the LASSO. Using the Hedge algorithm, we show that a our simple selection rule can achieve prediction results comparable to Cross Validation at a potentially much lower computational cost.

stat.CO↗

arXiv2017

Multiple Changepoint Estimation in High-Dimensional Gaussian Graphical Models

We consider the consistency properties of a regularised estimator for the simultaneous identification of both changepoints and graphical dependency structure in multivariate time-series. Traditionally, estimation of Gaussian Graphical Models (GGM) is performed in an i.i.d setting. More recently, such models have been extended to allow for changes in the distribution, but only where changepoints are known a-priori. In this work, we study the Group-Fused Graphical Lasso (GFGL) which penalises partial-correlations with an L1 penalty while simultaneously inducing block-wise smoothness over time to detect multiple changepoints. We present a proof of consistency for the estimator, both in terms of changepoints, and the structure of the graphical models in each segment.

math.ST↗

arXiv2015

Change-Point Estimation in High-Dimensional Markov Random Field Models

This paper investigates a change-point estimation problem in the context of high-dimensional Markov Random Field models. Change-points represent a key feature in many dynamically evolving network structures. The change-point estimate is obtained by maximizing a profile penalized pseudo-likelihood function under a sparsity assumption. We also derive a tight bound for the estimate, up to a logarithmic factor, even in settings where the number of possible edges in the network far exceeds the sample size. The performance of the proposed estimator is evaluated on synthetic data sets and is also used to explore voting patterns in the US Senate in the 1979-2012 period.

stat.ME↗

arXiv2014

Bayesian Inference in Nonparametric Dynamic State-Space Models

We introduce state-space models where the functionals of the observational and the evolutionary equations are unknown, and treated as random functions evolving with time. Thus, our model is nonparametric and generalizes the traditional parametric state-space models. This random function approach also frees us from the restrictive assumption that the functional forms, although time-dependent, are of fixed forms. The traditional approach of assuming known, parametric functional forms is questionable, particularly in state-space models, since the validation of the assumptions require data on both the observed time series and the latent states; however, data on the latter are not available in state-space models. We specify Gaussian processes as priors of the random functions and exploit the "look-up table approach" of \ctn{Bhattacharya07} to efficiently handle the dynamic structure of the model. We consider both univariate and multivariate situations, using the Markov chain Monte Carlo (MCMC) approach for studying the posterior distributions of interest. In the case of challenging multivariate situations we demonstrate that the newly developed Transformation-based MCMC (TMCMC) of \ctn{Dutta11} provides interesting and efficient alternatives to the usual proposal distributions. We illustrate our methods with a challenging multivariate simulated data set, where the true observational and the evolutionary equations are highly non-linear, and treated as unknown. The results we obtain are quite encouraging. Moreover, using our Gaussian process approach we analysed a real data set, which has also been analysed by \ctn{Shumway82} and \ctn{Carlin92} using the linearity assumption. Our analyses show that towards the end of the time series, the linearity assumption of the previous authors breaks down.

stat.ME↗