Events

Dec 1 - Dec 8, 2024

• 

  Improving Optimization of Likelihood-based Generative Models with One Line of Code

  Maurizio Filippone, Associate Professor, Statistics

  Dec 5, 12:00 - 13:00

  B9, L2, R2325

  Generative Models (GMs) have attracted considerable attention due to their tremendous success in various domains, such as computer vision where they are capable to generate impressive realistic-looking images.

Nov 24 - Dec 1, 2024

• 

  LoCoDL: Communication-Efficient Distributed Optimization with Local Training and Compression

  Laurent Condat, Senior Research Scientist, Computer, Electrical and Mathematical Sciences and Engineering

  Nov 27, 12:00 - 13:00

  B9, L3, R3125

  In distributed optimization, and even more in federated learning, communication is the main bottleneck. We introduce LoCoDL, a communication-efficient algorithm that leverages the two techniques of Local training, which reduces the communication frequency, and Compression with a large class of unbiased compressors that includes sparsification and quantization strategies.
• 

  Spatio-Temporal Statistics in Geo-Environmental Data Science

  Ying Sun, Associate Professor, Statistics

  Nov 26, 15:00 - 16:30

  B9, L2, R2322

  spatio-temporal statistics

  In this talk, I will discuss the contributions and ongoing research of my Environmental Statistics Research Group in the area of spatio-temporal statistics, with a particular focus on geo-environmental data science. Our work is primarily centered around the development and application of sophisticated statistical models that improve the understanding and management of environmental data characterized by their spatial and temporal variability. My group has made significant advances in developing better spatio-temporal models that effectively capture the complexities inherent in environmental datasets, as well as developing innovative software tools such as ExaGeoStat, ParallelVecchiaGP, and DeepKriging, which support the analysis of large-scale geostatistical datasets. During this presentation, I will also showcase our research contributions motivated by environmental applications, including multivariate time series visualization and clustering, panel data analysis for functional and spatial data, and statistical process monitoring.

Nov 17 - Nov 24, 2024

• 

  The KAUST 2024 Workshop on Statistics

  Nov 18, 09:00 - Nov 21, 17:00

  Auditorium between Building 4 - 5

  The KAUST Statistics Workshop will feature the latest research on statistical methods and modeling to address real-world challenges in health, environment, climate, energy and beyond.

May 26 - Jun 2, 2024

• 

  Deep Neural Networks for Large-scale Complex Spatial and Spatio-temporal Processes

  Pratik Nag , Ph.D. Student, Statistics

  May 29, 09:00 - 11:30

  B4 L5 R5220

  Environmental statistics play a critical role in various interconnected domains, encompassing weather and climate forecasting, air quality monitoring, and sustainable urban planning. However, because of their high inherent unpredictability and nonstationarity, modeling complex spatio-temporal dynamics of environmental processes is challenging. This dissertation develops a set of DNN based methods for large-scale spatial and spatio-temporal processes.

May 19 - May 26, 2024

• 

  Visualization, Characterization, and Forecasting of Multivariate and Functional Time Series

  Cristian Felipe Jiménez Varón, Ph.D. Student, Statistics

  May 22, 14:00 - 16:00

  B4 L5 R5220

  This thesis explores advanced statistical models and methods for analyzing multivariate and functional time series data. It focuses on various aspects of statistical analysis, including visualization, robust outlier detection, inference, and forecasting. It addresses challenges in outlier detection for functional data, quantile spectral estimation for multivariate time series, and high-dimensional functional time series forecasting, with applications in environmental, financial, and demographic fields.

May 5 - May 12, 2024

• 

  Intracellular "in silico microscopes" - Fully 3D Spatio-Temporal Virus Replication Model Simulations

  Gabriel Wittum, Professor, Applied Mathematics and Computational Sciences

  May 9, 12:00 - 13:00

  B9 L2 H2 H2

  Despite being small and simple structured in comparison to their victims, virus particles have the potential to harm severly and even kill highly developed species such as humans. To face upcoming virus pandemics, detailed quantitative biophysical understanding of intracellular virus replication mechanisms is crucial.

Apr 28 - May 5, 2024

• 

  Statistical Deep-Learning for Spatiotemporal Extremes

  Raphaël Huser, Associate Professor, Statistics

  May 2, 12:00 - 13:00

  B9 L2 H2 H2

  Rare, low-probability events often lead to the biggest impacts. Therefore, the development of statistical approaches for modeling, predicting and quantifying environmental risks associated with natural hazards is of utmost importance. In this seminar, I will show how statistical deep-learning methods can help solve challenges that arise when modeling complex and massive spatiotemporal extremes data.
• 

  Statistical Analysis of Topological Patterns in Dependence Networks of Brain Time Series Data

  Anass B. El-Yaagoubi, Ph.D. (former), Statistics

  May 2, 11:00 - 12:00

  B1 L4 R4102

  Given the complex nature of brain signals and the challenges involved in estimating its dependence and analyzing the emerging topological patterns, this dissertation introduces innovative statistical tools designed to explore both the functional and effective connectivity within brain networks. It sheds light on frequency-specific patterns in ADHD subjects and introduces a novel approach for examining the hierarchical structure of brain regions during seizures. Our work provides a novel perspective on the organization of brain networks and presents insight into how various conditions influence their complex structure.

Apr 21 - Apr 28, 2024

• 

  On Distribution-Free Risk Control

  Michael Jordan, Professor Emeritus, University of California, Berkeley

  Apr 24, 15:00 - 16:00

  B9 L4 R4225

  We introduce a framework for calibrating machine learning models so that their predictions satisfy explicit, finite-sample statistical guarantees. Our calibration algorithms work with any underlying model and (unknown) data-generating distribution and do not require model refitting.
• 

  An Alternative View on AI: Collaborative Learning, Statistical Incentives, and Social Welfare

  Michael Jordan, Professor Emeritus, University of California, Berkeley

  Apr 23, 12:00 - 13:00

  Auditorium between B2 and 3

  Artificial intelligence (AI) has focused on a paradigm in which intelligence inheres in a single, autonomous agent. Social issues are entirely secondary in this paradigm. When AI systems are deployed in social contexts, however, the overall design of such systems is often naive --- a centralized entity provides services to passive agents and reaps the rewards. Such a paradigm need not be the dominant paradigm for information technology. In a broader framing, agents are active, they are cooperative, and they wish to obtain value from their participation in learning-based systems. Agents may supply data and other resources to the system, only if it is in their interest to do so. Critically, intelligence inheres as much in the overall system as it does in individual agents, be they humans or computers. This is a perspective that is familiar in the social sciences, and a key theme in my work is that of bringing economics into contact with foundational issues in computing and data sciences. I'll emphasize some of the mathematical challenges that arise at this tripartite interface.

Apr 14 - Apr 21, 2024

• 

  Traveling wave solutions of Reaction-Diffusion equation in Biological mathematics

  Hoyoun Kim, Postdoctoral Research Fellow, Applied Mathematics and Computational Sciences

  Apr 18, 12:00 - 13:00

  B9 L2 H2 H2

  Traveling wave solutions of reaction-diffusion systems have been studied to explain wave propagation phenomena in biological organisms.
• 

  Towards Self-explainable Deep Learning Models

  Prof. Michael Kampffmeyer, UiT The Arctic University of Norway

  Apr 16, 16:30 - 17:00

  B1 L4 R 4102

  Despite the significant advancements deep learning models have brought to solving complex problems in the real world, their lack of transparency remains a significant barrier, particularly in deploying them within safety-critical contexts.
• 

  Can neural ODEs forecast weather?

  Dr. Markus Heinonen, Academy Research Fellow, Aalto Univeristy, Finland

  Apr 16, 16:00 - 16:30

  B1 L4 R 4102

  Neural ODEs have surfaced in the last decade as a new perspective on modelling dynamics by learning the time-derivative that drives the system evolution forward as a neural network.

Mar 31 - Apr 7, 2024

• 

  A Correlated Three-Dimensional Integral Fractional Ornstein-Uhlenbeck Process Model for Animal Movement

  Jose Hermenegildo Ramirez Gonzalez, Postdoctoral Research Fellow, Statistics

  Apr 4, 12:00 - 13:00

  B9 L2 H2 H2

  Predicting the paths of animals poses a significant challenge, given the intricate nature of their behaviors, the impact of unpredictable environmental elements, individual differences, and the scarcity of precise data on their movements.

Mar 24 - Mar 31, 2024

• 

  Classification of time series in high dimensions

  Sarbojit Roy, Postdoctoral Research Fellow, Statistics

  Mar 28, 12:00 - 13:00

  B9 L2 H2 H2

  As more and more modern time series data sets are becoming high dimensional, the problem of classification in this context has received increasing attention. We propose a statistical framework for classifying multivariate stationary Gaussian time series where the number of covariates, the length of the series, and the sample size, all grow to infinity.

Mar 17 - Mar 24, 2024

• 

  Tracking rare events within the ensemble Kalman filtering

  Gaukhar Shaimerdenova, Ph.D. Student, Applied Mathematics and Computational Sciences

  Mar 21, 12:00 - 13:00

  B9 L2 H2 H2

  In this work, we employ importance sampling (IS) techniques to track a small over-threshold probability of a running maximum associated with the solution of a stochastic differential equation (SDE) within the framework of ensemble Kalman filtering (EnKF).

Mar 3 - Mar 10, 2024

• 

  Acoustics with a spin: A new paradigm for acoustic metamaterials, scattering, and propagation of sound

  Mohamed D. Farhat, Research Scientist, Waves in Complex Media

  Mar 7, 12:00 - 13:00

  B9 L2 H2 H2

  When pressure acoustic waves interact with rotating scatterers, they undergo peculiar and intriguing characteristics. In this talk, I will discuss our recent findings on the physics of acoustic scattering and propagation in spinning fluids.

Feb 25 - Mar 3, 2024

• 

  Quadratic Forms in Gaussian Random Variables: A Survey of Theoretical Results and Applications

  Mohanad Ahmed, Research Specialist, Information Science Lab

  Feb 29, 12:00 - 13:00

  B9 L2 H2 H2

  Quadratic Forms in Gaussian Random Variables (QFGRV) appear in a wide variety of applications in signal processing, communications, performance and reliability engineering, genetics, etc.
• 

  Principled Scaling of Deep Neural Networks

  Soufiane Hayou, Postdoc, Simons Institute, UC Berkeley

  Feb 26, 09:00 - 10:00

  B9 L4 R4225

  Neural Networks

  Neural networks have achieved impressive performance in many applications such as image and speech recognition and generation. State-of-the-art performance is usually achieved via a series of engineered modifications to existing neural architectures and their training procedures. However, a common feature of these systems is their large-scale nature: modern neural networks usually contain Billions - if not 10's of Billions - of trainable parameters, and empirical evaluations (generally) support the claim that increasing the scale of neural networks (e.g. width and depth) boosts the model performance if done correctly. However, given a neural network model, it is not straightforward to address the crucial question `how do we scale the network?'. In this talk, I will show how we can leverage different mathematical results to efficiently scale neural networks, with empirically confirmed benefits.

Feb 11 - Feb 18, 2024

• 

  Fuzzy clustering of circular time series based on a new distance with applications to wind data

  Ángel López Oriona, Postdoctoral Research Fellow, Statistics

  Feb 15, 12:00 - 13:00

  B9 L2 H2 H2

  Time series clustering is an essential machine learning task with applications in many disciplines. While the majority of the methods focus on time series taking values on the real line, very few works consider time series defined on the unit circle, although the latter objects frequently arise in many applications. In this talk, the problem of clustering circular time series is discussed.

Feb 4 - Feb 11, 2024

• 

  Dependence and Causality in Functional Brain Networks

  Hernando Ombao, Professor, Statistics

  Feb 8, 12:00 - 13:10

  B9 L2 H2 H2

  Brain activity over the entire network is complex. A full understanding of brain activity requires careful study of its multi-scale spatial-temporal organization. Motivated by these challenges, we will explore some characterizations of dependence between components of a multivariate time series and then apply these to the study of brain functional connectivity.

Jan 28 - Feb 4, 2024

• 

  Think globally, act locally: from domain decomposition to nonlinear model order reduction

  Prof. Charbel Farhat

  Jan 30, 14:00 - 15:00

  KAUST Auditorium R2030

  Abstract The Finite Element Tearing and Interconnecting (FETI) methodology is a family of domain decomposition (DD)-based iterative methods for the fast solution on massively parallel processors of large-scale systems of equations arising from finite element (FE) spatial approximations. Developed in its basic form in 1991, it has been extended and refined during 18 years to enable ever faster and more accurate analyses of a broad range of engineering and pathfinder systems such as re-entry vehicles, modern aircraft, advanced Formula 1 cars, aircraft carriers, underwater systems, electronic

Jan 21 - Jan 28, 2024

• 

  On space-time models: models and applications

  Elias Teixeira Krainski, Postdoctoral Research Fellow, Statistics

  Jan 25, 12:00 - 13:00

  B9 L2 H2 H2

  Partial differential equations are a mathematical tool widely used to model phenomena in several different fields. A stochastic partial differential equation (SPDE) introduces random forcing to take the nature of real-world observations.

Dec 3 - Dec 10, 2023

• 

  Multilevel Monte Carlo for a class of Partially Observed Processes in Neuroscience

  Dec 7, 12:00 - 13:00

  B9 L2 H2

  In this talk, we consider Bayesian parameter inference associated to a class of partially observed stochastic differential equations (SDE) driven by jump processes. Such type of models can be routinely found in applications, of which we focus on the case of neuroscience.

Nov 26 - Dec 3, 2023

• 

  Discretization and Optimization in Applied Geometry

  Nov 30, 12:00 - 13:00

  B9 L2 H2

  Many problems in applied geometry amount to the solution of a typically nonlinear partial differential equation. We will discuss why it may not be a good idea to discretize the equation, but to take the viewpoint of discrete differential geometry and discretize the theory.

Nov 19 - Nov 26, 2023

• 

  Efficient Inference for Joint Models of Multivariate Longitudinal and Survival Data Using INLAjoint

  Denis Rustand, Postdoctoral Research Fellow, Statistics

  Nov 23, 12:00 - 13:00

  B9 L2 H2

  Clinical research often requires the simultaneous study of longitudinal repeated measurements and time-to-event (i.e., survival) data. Joint models, which can combine these two types of data, are invaluable tools in this context.

Nov 12 - Nov 19, 2023

• 

  How do periodic structures affect nonlinear waves?

  David Ketcheson, Professor, Applied Mathematics and Computational Sciences

  Nov 16, 12:00 - 13:00

  B9 L2 H2

  We are all familiar with the tendency of water waves to break in shallow water, for instance at the beach. Indeed, breaking is a universal behavior of solutions to first-order nonlinear hyperbolic PDEs, and manifests itself in phenomena ranging from traffic jams to shock waves.
• 

  Experimental Design Statistical Workshop

  Hernando Ombao, Professor, Statistics

  Nov 14, 09:00 - Nov 15, 17:00

  B1 L4 R4102

  This workshop will cover the statistical essentials to designing experiments including power analysis, sample size calculation, regression, and variance analysis. A statistically sound experimental design is essential in applying for research grants. Detailed syllabi will be sent to those who register. Limited seats. Sign up now!

Nov 5 - Nov 19, 2023

• 

  Family of Mean-Mixtures of Multivariate Normal Distributions

  Prof. Narayanaswamy Balakrishnan, Department of Mathematics and Statistics, McMaster University

  Nov 12, 15:30 - 16:30

  B1 L4 R4102

  In this talk, I will describe the family of mean-mixtures of multivariate normal distributions and establish many of its properties, stochastic representations, moments, distributional shape characteristics, etc.
• 

  Models and Inference for Non-Gaussian Random Vectors and Fields

  Sagnik Mondal, Ph.D. Student, Statistics

  Nov 12, 12:30 - 14:30

  B5 L5 R5209

  The multivariate Gaussian distribution is widely used in many statistical applications due to its appealing features. However, real-world data often violate its assumptions, showing skewness and/or tail-thickness.

Nov 5 - Nov 12, 2023

• 

  The KAUST 2023 Workshop on Statistics

  Nov 6, 09:00 - Nov 9, 17:00

  B18 R4203

  statistics data science modeling big data sustainability health environment

  The workshop will feature the latest research on statistical methods and modeling to address real-world challenges in health, environment, and sustainability.

Oct 29 - Nov 5, 2023

• 

  Free Boundary Problems in Science and Engineering

  Rafayel Teymurazyan, Research Scientist, Applied Mathematics and Computational Sciences

  Nov 2, 12:00 - 13:00

  B9 L2 H2

  Free boundary problems arise naturally in a range of mathematical models that describe physical, biological or financial phenomena, such as the melting of ice into water, the dynamics of a population or the behavior of stock markets, to mention just a few.
• 

  Spatial and Spatio-temporal Statistical Methods for Environment and Public Health Applications

  Andre Victor Ribeiro Amaral , Ph.D. Student, Statistics

  Nov 2, 11:45 - 14:30

  B9 L4 R4225

  This thesis proposes statistical spatial and spatial-temporal methods for addressing real-world challenges within the public health and environmental domains.

Oct 22 - Oct 29, 2023

• 

  Spatio-temporal Bayesian analysis of excess mortality in 5 European countries in 2020

  Prof. Virgilio Gómez-Rubio, Department of Mathematics, Universidad de Castilla-La Mancha, Spain

  Oct 26, 12:00 - 13:00

  B9 L2 H2

  Abstract The COVID-19 pandemic produced excess mortality in many countries in 2020. In my talk, I will illustrate our use of Bayesian models and R software for the analysis of total deaths and population data, combined with environmental data, to the study of excess mortality in 2020. The study analyzes total deaths by age groups and sex in 5 European countries, and provides estimates of excess mortality at different levels of aggregation (province, autonomous community and country). For both data analysis and visualization of the results we have used the statistical software R. In addition

Oct 15 - Oct 22, 2023

• 

  Latent Gaussian Spatial Modelling with Change of Support

  Erick Chacon Montalvan, Postdoctoral fellow, Statistics Geohealth Group, KAUST

  Oct 19, 12:00 - 13:00

  B9 L2 H2

  Spatial data analysis commonly needs to deal with spatial data derived from multiple sources (e.g. satellites, stations, survey samples) with different supports, but associated with the same properties of a spatial phenomenon under interest. Usually, predictors are also measured on different spatial supports than the response variable.

Oct 8 - Oct 15, 2023

• 

  Space–Time Multiscale Modeling and Numerics of Sorption Kinetics

  Clarissa Astuto, Postdoctoral Research Fellow, Applied Mathematics and Computational Sciences

  Oct 12, 12:00 - 13:00

  B9 L2 H2

  In this talk we propose and validate a Space Multiscale model for the description of particle diffusion in the presence of trapping boundaries. We start from a drift diffusion equation in which the drift term describes the effect of bubble traps, and it is simulated by the Lennard–Jones potential.

Oct 1 - Oct 8, 2023

• 

  How many sample points do we need for least squares?

  Diogo Gomes, Program Chair, Applied Mathematics and Computational Sciences

  Oct 5, 12:00 - 13:00

  B9 L2 H2

  Abstract The goal of the least squares method is to find the best linear combination of certain given functions to fit given data. A typical example is the following: given pairs $(x_i, y_i)$, find the coefficients, $a$ and $b$ that minimize the error $\sum_i (y_i-a -b x_i)^2$. In this case, if at least two of the coordinates $x_i$ are distinct, we can always find coefficients $a$ and $b$. However, if the $x_i$ are randomly sampled, these coordinates may be extremely close, creating large errors in the estimated values of $a$ and $b$. In this talk, we investigate lower bounds on the minimum

Sep 24 - Oct 1, 2023

• 

  Decision Trees for Fault Diagnosis in Circuits and Switching Networks - 2023-09-28

  Mikhail Moshkov, Professor, Applied Mathematics and Computational Sciences

  Sep 28, 12:00 - 13:00

  B9 L2 H2

  We study theoretical problems of fault diagnosis in circuits and switching networks, which are among the most fundamental models for computing Boolean functions.

Sep 17 - Sep 24, 2023

• 

  Predictive Performance Test based on the Exhaustive Nested Cross-Validation for High-dimensional data

  Iris Ivy Gauran, Postdoctoral Research Fellow, Biostatistics Group

  Sep 21, 12:00 - 13:00

  B9 L2 H2

  Cross-validation is an algorithmic technique extensively used for estimating the prediction error, tuning the regularization parameter, and choosing between competing predictive rules.

Sep 10 - Sep 17, 2023

• 

  Goodness-of-fit tests for multivariate skewed distributions based on the characteristic function

  Maicon J. Karling, Postdoctoral Research Fellow, Statistics

  Sep 14, 12:00 - 13:00

  B9 L2 H2

  Goodness-of-fit tests determine how well a set of observed data fits a particular probability distribution. They can also show if some categorical variable follows a hypothesized family of distributions.
• 

  Spatial Models and Extreme-Value Methods for Wildfire Risk Assessment

  Daniela Cisneros, Ph.D. Student, Statistics

  Sep 11, 16:00 - 17:00

  B3 L5 R5220

  extreme statistics Applied Machine Learning geospatial statistics

  The statistical modeling of spatial and extreme events provides a framework for the development of techniques and models to describe natural phenomena in a variety of environmental, geoscience, and climate science applications. In a changing climate, various natural hazards, such as wildfires, are believed to have evolved in frequency, size, and spatial extent, although regional responses may vary.

Sep 3 - Sep 10, 2023

• 

  From Euler flows with friction to gradient flows and applications

  Athanasios Tzavaras, Professor, Applied Mathematics and Computational Sciences

  Sep 7, 12:00 - 13:00

  B9 L2 H2

  I will review some works on the high-friction limit (or small mass approximation) from Euler flows to advection-diffusion systems that are gradient flows, and related asymptotic problems in fluid mechanics. The formulation exploits the variational structure of compressible Euler flows and is connected to the interpretation of nonlinear Fokker-Planck systems as gradient flows in Wasserstein distance.

Aug 27 - Sep 3, 2023

• 

  Efficient Intensity Estimation Using Adaptive Bandwidths on Linear Networks: A Study and Application in Traffic Accident Analysis

  Jonatan A. González, Postdoctoral Research Fellow, Statistics

  Aug 31, 12:00 - 13:00

  B9 L2 H2

  Estimating first-order intensity functions is crucial in the analysis of point patterns on linear networks, but selecting suitable bandwidths for non-parametric methods remains challenging. We propose an adaptive intensity estimator for the heating kernel that adjusts bandwidths based on data points, a novel approach in this context.

May 28 - Jun 11, 2023

• 

  Approximate Bayesian inference based on dense matrices and new features using INLA

  Jun 4, 15:00 - 16:00

  B4 L5 R5220

  Bayesian computational statistics

  The Integrated Nested Laplace Approximations (INLA) method has become a commonly used tool for researchers and practitioners to perform approximate Bayesian inference for various fields of applications. It has become essential to incorporate more complex models and expand the method’s capabilities with more features. In this dissertation, we contribute to the INLA method in different aspects.
• 

  Stochastic environmental modeling in a time of convergence: physics meets artificial intelligence

  Prof. Stefano Castruccio, Associate professor, University of Notre Dame, USA

  Jun 4, 10:00 - 11:00

  B1 L4 R4102

  It is widely acknowledged how the relentless surge of Volume, Velocity and Variety of data, as well as the simultaneous increase of computational resources have stimulated the development of data-driven methods with unprecedented flexibility and predictive power. However, not every environmental study entails a large data set: many applications ranging from astronomy or paleo-climatology have a high associated sampling cost and are instead constrained by physics-informed partial differential equations. Throughout the past few years, a new and powerful paradigm has emerged in the machine learning literature, merging data-driven and physics-informed problems, hence providing a unified framework for a whole spectrum of problems ranging from data-rich/context-poor to data-poor/context-rich. In this talk, I will present this new framework and discuss some of the most recent efforts to reformulate it as a stochastic model-based approach, thereby allowing calibrated uncertainty quantification.

May 28 - Jun 4, 2023

• 

  Leave-Group-Out Cross-Validation for Latent Gaussian Models

  Zhedong Liu, Ph.D. Student, Statistics

  May 30, 15:30 - 17:30

  B1 R4102;

  The commonly used leave-one-out and K-fold cross-validation methods are not suitable for structured models with multiple prediction tasks. To overcome this limitation, we introduce leave-group-out cross-validation, which allows groups to adapt to different tasks. We propose an automatic group construction method and provide an efficient approximation for latent Gaussian models. Moreover, this method is conveniently implemented in the R-INLA software.

May 21 - Jun 4, 2023

• 

  Criticism and robustification of latent Gaussian models

  Rafael Medeiros Cabral, Ph.D. Student, Statistics

  May 28, 15:00 - 16:00

  B1 L4 R4102

  latent Gaussian models

  Latent Gaussian models (LGM) are widely used but struggle with certain datasets that contain non-Gaussian features, such as sudden jumps or spikes. This dissertation aims to provide tools for researchers to check the adequacy of the fitted LGM (criticism); if the check fails, offer efficient and user-friendly implementations of latent non-Gaussian models, which lead to more robust inferences (robustification).

May 7 - May 14, 2023

• 

  New Graphical Displays for Classification

  Peter Rousseeuw, Professor Emeritus, Statistics and Data Science, KU Leuven, Belgium

  May 9, 15:00 - 16:00

  B9 L2 H2

  Classification is a major tool of statistics and machine learning. Several classifiers have interesting visualizations of their inner workings. Here we pursue a different goal, which is to visualize the cases being classified, either in training data or in test data. An important aspect is whether a case has been classified to its given class (label) or whether the classifier wants to assign it to a different class. This is reflected in the probability of the alternative class (PAC). A high PAC indicates label bias, i.e. the possibility that the case was mislabeled. The PAC is used to construct a silhouette plot which is similar in spirit to the silhouette plot for cluster analysis. The average silhouette width can be used to compare different classifications of the same dataset. We will also draw quasi residual plots of the PAC versus a data feature, which may lead to more insight in the data. One of these data features is how far each case lies from its given class, yielding so-called class maps. The proposed displays are constructed for discriminant analysis, k-nearest neighbors, support vector machines, CART, random forests, and neural networks. The graphical displays are illustrated and interpreted on data sets containing images, mixed features, and texts.
• 

  Detecting Cellwise Outliers in Your Data

  Peter Rousseeuw, Professor Emeritus, Statistics and Data Science, KU Leuven, Belgium

  May 9, 12:00 - 13:00

  B9 L2 H2

  A multivariate dataset consists of n cases in d dimensions, and is often stored in an n by d data matrix. It is well-known that real data may contain outliers. Depending on the situation, outliers may be (a) undesirable errors which can adversely affect the data analysis, or (b) valuable nuggets of unexpected information. In statistics and data analysis the word outlier usually refers to a row of the data matrix, and the methods to detect such outliers only work when at least half the rows are clean. But often many rows have a few contaminated cell values, which may not be visible by looking at each variable (column) separately. We describe a method to detect deviating data cells in a multivariate sample which takes the correlations between the variables into account. It has no restriction on the number of clean rows, and can deal with high dimensions. Other advantages are that it provides predicted values of the outlying cells, while imputing missing values at the same time. We illustrate the method on several real data sets, where it uncovers more structure than found by purely columnwise methods or purely rowwise methods. The proposed method can help to diagnose why a certain row is outlying, e.g. in process control. It also serves as an initial step for estimating multivariate location and scatter matrices, and for cellwise robust principal component analysis.