Presentation Abstract

Department of Science and Mathematics at TexasA&M University-Central Texas.

Title:  How to Become an Effective Practitioner of Machine Learning.

Abstract: Machine learning (including deep learning) has become a cornerstone of modern technology.  The international job market is wide open for people who understand machine learning and can use it effectively. However, machine learning is not easy, and many people are jumping into the field without the necessary skills and background knowledge. In this talk, I will introduce free learning resources (mostly online) and practices that will enable you to master machine learning principles and gain hands-on experience, which will enable you to become truly proficient. 

School of Mathematical Sciences at Rochester Institute of Technology, USA.

Title: A Tour of Classical and Emerging Tools for Modern Data Era Statistical Model Building.

 

Abstract:  At its core, practical statistical model building is ideally approached as a suitable combination of art and science, indeed of intuition and extuition, of commonsense reasoning coupled with elaborate technical schemes. This presentation intends to provide a gentle and accessible, yet rigorous tour of some of the most frequently used methods, techniques, tools and algorithms currently prevalent in statistical model building. I will use the well known framework of statistical regression analysis as an anchor from which to stretch and touch on aspects of parameter estimation and inference and model selection using various criteria. I intend to highlight both classical and Bayesian views, but also bring in more predictively minded paradigms like the ensemble learning method of random forest born from classification and regression trees. Along with ensemble learning, I will finally touch on the modern paradigm of regularization learning with a focus on Generalized Linear Model Networks (GLMNET) and its great appeal in the now ubiquitous large p small n data scenario. Last be not least, I will very briefly evoke kernel regression and deep neural networks in the context of supervised learning. Wherever possible and/or deemed necessary, I will provide pointers to practical implementations in the R statistical software environment.

RHODES University, South Africa.

Title: Perspectives on Long Term Research Data Archiving Solutions: the case of the SKA with learning algorithms.

 

Abstract: This talk will address the data format and archiving issues for the Square Kilometre Array. One of the major contributors to the large data volumes is the maximum baseline of an array since it sets how well the data needs to be sampled to avoid smearing and decorrelation of the signal. Now, the sampling rate is dependent on the baseline length as shorter baselines can be sampled a lot more coarsely which leads to smaller data volumes. In fact, baseline dependent averaging (BDA) is an established technique for compressing radio interferometers, however, this technique results in irregularly sampled data which while supported by the database format requires the data to be restructured in ways that reduce performance when processing. But more importantly, radio-astronomy software tools implicitly assume uniformly sampled datasets. Recent research in (M. Atemkeng, in prep). in collaboration with the Wits University and SARAO; shows an approach that uses local  low-rank matrix approximation to achieve similar compression rates while significantly minimizing smearing. At the end of this talk, I will discuss why Central African countries should join the SKA; a mega science project.

Steam, USA.

Title: AI For Remote Collaboration via Augmented Reality.

Abstract: Remote Augmented Reality unlocks a new set of experiences and use cases for customers. One use case in particular is remote collaboration where an expert in a different location can help an onsite customer with a task in the customer environment (for example repairing an appliance). This is critical in a Digital Economy where the workforce is becoming more distributed geographically, expert knowledge is a soughtafter asset, and more consumers are equipped with smartphones. Streem focuses precisely on the mobile technology that unlocks this and enables improved customer experiences with AR-supported video tools. This talk will dive into how Remote AR is used to create a new type of customer engagement, and the Artificial Intelligence solutions that support these experiences.

Université de Douala, Cameroun.

Title: Méthode d’authentification continue des intrusions dans un système informatique à l’aide des outils de la Théorie des jeux non coopératifs.

Abstract : L’objectif de ce travail est l’analyse des mécanismes d’authentification à travers l’approche de la théorie des jeux. Plus précisément, nous formulons l’interaction entre l’attaquant et le défenseur comme un jeu dynamique stochastique de type learderfollower. Nous fournissons ensuite une caractérisation de la stratégie d’attaque optimale, et montrons qu’elle possède une structure de seuil. Concernant l’optimisation du paramètre du défenseur, nous fournissons une caractérisation de l’impact du paramètre d’authentification continue sur les utilités de l’attaquant et du défenseur. Enfin, nous fournissons des résultats numériques pour illustrer la stratégie de l’attaquant et l’impact de la stratégie du défenseur sur les utilités attendues de l’attaquant et du défenseur.