A5091752605- Point processes and geometric inequalities
- Markov Chains and Monte Carlo Methods
- Statistical Methods and Inference
- Probability and Risk Models
- Advanced Queuing Theory Analysis
- Simulation Techniques and Applications
- Bayesian Methods and Mixture Models
- Stochastic processes and statistical mechanics
- Machine Learning and Data Classification
- Neural Networks and Applications
- Fault Detection and Control Systems
- Gaussian Processes and Bayesian Inference
- Advanced Control Systems Optimization
- Sparse and Compressive Sensing Techniques
- Multi-Criteria Decision Making
- COVID-19 epidemiological studies
- Stochastic Gradient Optimization Techniques
- Statistical and numerical algorithms
- Network Traffic and Congestion Control
- Statistical Methods and Bayesian Inference
- COVID-19 Digital Contact Tracing
- Wireless Networks and Protocols
- Advanced Bandit Algorithms Research
- Computational Physics and Python Applications
- Advanced Data Processing Techniques
UNSW Sydney
2023-2024
Macquarie University
2021-2024
The University of Queensland
2017-2021
Tata Institute of Fundamental Research
2013-2015
Dallas County
2013
Chromosome arm aneuploidies (CAAs) are pervasive in cancers. However, how they affect cancer development, prognosis and treatment remains largely unknown. Here, we analyse CAA profiles of 23,427 tumours, identifying aspects tumour evolution including probable orders which CAAs occur predicting tissue-specific metastasis. Both haematological solid cancers initially gain chromosome arms, while only subsequently preferentially lose multiple arms. 72 88 synergistically co-occurring pairs...
We consider the multi-armed restless bandit problem (RMABP) with an infinite horizon average cost objective. Each arm of RMABP is associated a Markov process that operates in two modes: active and passive. At each time slot controller needs to designate subset arms be active, which processes will evolve differently from passive case. Treated as optimal control problem, solution known computationally intractable. In many cases, Whittle index policy achieves near performance can tractably...
Abstract How do fine modifications to social distancing measures really affect COVID-19 spread? A major problem for health authorities is that we not know. In an imaginary world, might develop a harmless biological virus spreads just like COVID-19, but traceable via cheap and reliable diagnosis. By introducing such into the population observing how it spreads, would have way of learning about because benign respond behaviour in similar manner. Such does exist. Instead, propose safe...
Abstract The problem of best subset selection in linear regression is considered with the aim to find a fixed size features that fits response. This particularly challenging when total available number very large compared data samples. Existing optimal methods for solving this tend be slow while fast have low accuracy. Ideally, new perform faster than existing but comparable accuracy, or, being more accurate computational speed. Here, we propose novel continuous optimization method...
We consider spatial marked Poisson arrivals in a Polish space. These are accepted or lost general state dependent manner. The remain the system for random amount of time, where individual sojourn times i.i.d. For such systems, we develop semi-closed form expressions steady probabilities that can be seen to insensitive time distribution, and rely essentially on static objects meeting acceptance criteria. latter observation is then exploited yield straightforward exact simulation algorithms...
Multiclass open queueing networks find wide applications in communication, computer, and fabrication networks. Steady-state performance measures associated with these is often a topic of interset. Conceptually, under mild conditions, sequence regeneration times exists multiclass networks, making them amenable to regenerative simulation for estimating steady-state measures. However, typically, identification such difficult. A well-known exception when all interarrival are exponentially...
We propose a continuous optimization algorithm for the Column Subset Selection Problem (CSSP) and Nyström approximation. The CSSP method construct low-rank approximations of matrices based on predetermined subset columns. It is well known that choosing best column size k difficult combinatorial problem. In this work, we show how one can approximate optimal solution by defining penalized loss function minimized via stochastic gradient descent. gradients be estimated efficiently using...
In recent years, Monte Carlo estimators have been proposed that can estimate the ratio of two expectations without bias. We investigate theoretical properties a Taylor-expansion based estimator reciprocal mean non-negative random variable. establish explicit expressions for computational efficiency this and obtain optimal choices its parameters. also derive corresponding practical confidence intervals show they are asymptotically equivalent to maximum likelihood (biased) as simulation budget...
Conceptually, under restrictions, multiclass open queueing networks are positive Harris recurrent Markov processes, making them amenable to regenerative simulation for estimating the steady-state performance measures. However, regenerations in such difficult identify when interarrival times generally distributed. We assume that have exponential or heavier tails and show distributions can be decomposed into mixture of sums independent random variables at least one components is exponentially...
Compositional data find broad application across diverse fields due to their efficacy in representing proportions or percentages of various components within a whole. Spatial dependencies often exist compositional data, particularly when the represents different land uses ecological variables. Ignoring spatial autocorrelations modelling may lead incorrect estimates parameters. Hence, it is essential incorporate information into statistical analysis obtain accurate and reliable results....
The selection of best variables is a challenging problem in supervised and unsupervised learning, especially high dimensional contexts where the number usually much larger than observations. In this paper, we focus on two multivariate statistical methods: principal components analysis partial least squares. Both approaches are popular linear dimension-reduction methods with numerous applications several fields including genomics, biology, environmental science, engineering. particular, these...
We present a new optimization method for the group selection problem in linear regression. In this problem, predictors are assumed to have natural structure and goal is select small set of groups that best fits response. The incorporation predictor matrix key factor obtaining better estimators identifying associations between response predictors. Such discrete constrained well-known be hard, particularly high-dimensional settings where number much larger than observations. propose tackle by...
We propose a player rating mechanism for Counter-Strike: Global Offensive (CS ), popular e-sport, by analyzing players' Plus/Minus values. The value represents the average point difference between player's team and opponent's across all matches has participated in. Using models such as regularized linear regression, logistic Bayesian models, we examine relationship participation differences. most commonly used metric in CS community is "Rating 2.0," which focuses solely on individual...
ABSTRACT The selection of best variables is a challenging problem in supervised and unsupervised learning, especially high‐dimensional contexts where the number usually much larger than observations. In this paper, we focus on two multivariate statistical methods: principal components analysis partial least squares. Both approaches are popular linear dimension‐reduction methods with numerous applications several fields including genomics, biology, environmental science, engineering....