The 66th Bernard Price Memorial Lecture - Kwa-Zulu Natal Centre



The 66th Bernard Price Memorial Lecture - Kwa-Zulu Natal Centre
Data driven machine learning algorithms have become the preferred approaches for analysing complex societal challenges, addressing a vast array of problems while opening new avenues for technological development.
Event Name: The 66th Bernard Price Memorial Lecture - Kwa-Zulu Natal Centre
Presenter(s): Prof. Fulufhelo Nelwamondo
Event Date: 13 September 2017
Event Time: 17:00 - 20:00
Event Type: Annual Lecture
Event Theme: BP Lecture
Venue: Lecture Theatre, eThekweni Training Centre, 17 Supply Road, Springfield, Durban
Province: KZN
Cost:
RSVP From: 11 September 2017
RSVP to: 13 September 2017
Event Contact: Gill Nortier
saiee@iafrica.com
(c) 073-739-7284
To RSVP: Click here

The 66th Bernard Price Memorial Lecture - Kwa-Zulu Natal Centre





 

SOUTH AFRICAN INSTITUTE OF ELECTRICAL ENGINEERS

PRESENTS THE

66th Bernard Price Memorial Lecture

“REDRESSING STRUCTURAL AND SYSTEMIC BIAS IN

MODERN-DAY AUTOMATED SOLUTIONS”


presented by

Prof Fulufhelo Nelwamondo


13 September 2017


Lecture Theatre, eThekweni Training Centre

17 Supply Road, Springfield

Durban


17h00 for 17h30


~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~


THE 2017 BERNARD PRICE MEMORIAL LECTURER



Prof Fulufhelo Nelwamondo

Biography

Fulufhelo Nelwamondo is an electrical engineer by training, and holds a Bachelor of Science and a PhD in Electrical Engineering, in the area of Computational Intelligence, both from the University of the Witwatersrand, in South Africa. He is a registered Professional Engineer, a Member of the SAIEE, and a senior member of the IEEE. He is the Executive Director for the CSIR Modelling and Digital Science Unit and a visiting professor of Electrical Engineering at the University of Johannesburg. He previously was a post-doctoral fellow at the Graduate School of Arts and Sciences, of Harvard University.

Prof Nelwamondo has research and practical experience in software engineering, computational intelligence and optimisation in various applications. He is the youngest South African ever to receive the Harvard-South Africa fellowship and has been awarded many national and international research accolades, the latest being Order of Mapungubwe in Silver, which he received in 2017. Other accolades he received come from organizations such as the IEEE, SAIEE, National Science and Technology Forum, Springer, amongst others. Prof Nelwamondo has interests in exciting and emerging areas of software and technology applications including biometrics based system, data mining, modelling of complex systems, machine learning, optimisation and mechanism design. He has presented his work in various countries across the world and has published over 100 papers in reviewed journals, conferences and book chapters. Prof Nelwamondo has successfully supervised 20 Doctoral and Masters Degrees in Electrical Engineering and Computer Science.

 

66th Bernard Price Memorial Lecture

“Redressing structural and systemic bias in

modern-day automated solutions”

 

Data driven machine learning algorithms have become the preferred approaches for analysing complex societal challenges, addressing a vast array of problems while opening new avenues for technological development. These algorithms typically use large amounts of data to train architectures that identify underlying patterns in data, and then leverage these patterns to provide engineering solutions to complex problems, and allows for large scale automated decision making.

Unfortunately, these algorithms are susceptible to structural and systematic bias, resulting from data limitations, which often leads to unwanted discrimination. Some malevolent side effect of this automation phenomenon is the unwitting introduction of race and gender disparities as well as other prejudicial classifications. This bias potentially leads to undesired conclusions with discriminatory or segregatory connotations. Bias of this kind can be hard to detect and redress, particularly when the models exhibiting these properties contain millions of parameters with complex interactions of high dimensional data.

This lecture will present some of the challenges of embedded prejudice in machine learning algorithms and will cover some observed examples and the societal effects thereof. In addition, the lecture will present ways of designing automatic algorithms that can detect bias of this form and some transfer learning mechanisms that can correct problematic machine learning models.

To RSVP: Click here