Algorithms in the 4IR: Improved machine learning techniques via probability theory

Speaker

Dr Peter Olukanmi

Title

Algorithms in the 4IR: Improved machine learning techniques via probability theory

Link: https://zoom.us/j/96223068629?pwd=ajRKYnU0UEVER3h0cXJwWTJuTkdqZz09

Date: Tuesday 31 May 2022

Time: 14:00 (2:00PM).

Abstract

The big idea behind this talk is to inspire researchers and students in the mathematical sciences towards global impact in the current 4IR era, by harnessing their existing tools to design machine learning algorithms challenge currently well-celebrated algorithms. As a specific case study, this talk discusses the speakers work in addressing key drawbacks of one of the most widely used algorithms (k-means clustering). There is practically no modern ML software that does not implement k-means, and addressing its drawbacks has been of interest among top ML scientists. This talk discusses: (1) automatic estimation of k, the number of clusters, (2) completely automated simultaneous outlier detection and clustering (a reputedly difficult problem), (3) ultrafast (constant-time) clustering without accuracy loss. Interestingly, each of these breakthroughs is rooted in one of two well-known probability theory tools: the central limit theorem and Chebyshev’s inequality. Compared to other existing improvements, our algorithms are simple, explainable, and fast even on large datasets. These are crucial qualities algorithms must possess to be candidates for real-world use.

Short Bio

I lead the Data Science and Computational Intelligence Research Cluster (D-CIRCLE) as a Senior Lecturer in the Department of Electrical Engineering Technology (DEET) at the University of Johannesburg (UJ). I am also a member of the Johannesburg Machine Intelligence Lab (JMIL), a former postdoctoral fellow at UJ and a former Lecturer at the Sol Plaatje University. I obtained my PhD under the Institute for Intelligent Systems, at UJ; MSc in Computer Science from the University of KwaZulu-Natal (UKZN); and BSc Honors from the University of Lagos. I have taught several university modules such as Machine Learning, Signal & Image Processing, Advanced Analysis of Algorithms, and Simulation and Modelling. My interests span both fundamental and applied research in Data Science. I have a few top-quarter (Q1) journal articles and I won two IEEE conference awards. I like spending time with family and playing musical instruments.

Disclaimer: The University of Johannesburg encourages academic debate and discussion that are conducted in a manner that upholds respectful interaction, safety of all involved, and freedom of association as enshrined in the law, the Constitution, and within the boundaries of the University policies. The views expressed during events are expressed in a personal capacity and do not necessarily reflect the views of the University of Johannesburg.