Maties Machine Learning (MML) is a seminar series and discussion forum with the goal of bringing together people working on machine learning at Stellenbosch University. We meet roughly every second week for short talks on people’s current work, some ML-related topic, or open discussions. The idea is to get to know what others are working on and to strengthen machine learning research at Stellenbosch.

Upcoming talks

14 February 2020, 13:00-13:50 in K303 (Knowledge Centre, Engineering)

  • Adam Viktorin - Evolutionary Computation - Self-adaptive Differential Evolution

    A current trend in the area of heuristic optimization is the use of algorithms based on the Darwinian principle of evolution. Differential Evolution (DE) is just one of the many cases that follow this principle. Where other approaches are usually based on a metaphor from nature, DE is based on a simple engineering idea of vector computation. This might be one of the reasons why it is not overly popular among practitioners. Still, its excellent performance (proven over and over on numerous competitions) is a reason why it should not be overlooked. In this presentation, Adam will talk about his study in the recent field of self-adaptive DE, which softens the famous no free lunch theorem of optimization.

  • Michal Pluhacek - Swarm Intelligence - From Metaheuristics to Swarm Robotics

    Swarm Intelligence (SI) refers to the ability of animal collectives to exhibit surprisingly complex and intelligent behavior. Such behavior has been observed in insects, fish, birds and many other animal species. In the past decade, the term Swarm Intelligence was connected mostly with Artificial Intelligence (AI) research efforts. In the field of metaheuristic optimizers, there is a large group of algorithms that are inspired by swarm intelligence (or claim to be). More recently, the focus of the research community seems to be shifting more and more towards Swarm Robotics (SR), a dynamically growing field of research, where the swarm intelligence principles observed in nature and simulated in many algorithms are put into practice in controlling the behavior of decentralized and self-organizing robotic swarms.

17 February 2020, 13:00-13:50 in K302 (Knowledge Centre, Engineering)

Scott Cameron - Stochastic Gradient Annealed Importance Sampling for Efficient Online Marginal Likelihood Estimation

We consider estimating the marginal likelihood in settings with independent and identically distributed (i.i.d.) data. We propose estimating the predictive distributions in a sequential factorization of the marginal likelihood in such settings by using stochastic gradient Markov Chain Monte Carlo techniques. This approach is far more efficient than traditional marginal likelihood estimation techniques such as nested sampling and annealed importance sampling due to its use of mini-batches to approximate the likelihood. Stability of the estimates is provided by an adaptive annealing schedule. The resulting stochastic gradient annealed importance sampling (SGAIS) technique, which is the key contribution of our paper, enables us to estimate the marginal likelihood of a number of models considerably faster than traditional approaches, with no noticeable loss of accuracy. An important benefit of our approach is that the marginal likelihood is calculated in an online fashion as data becomes available, allowing the estimates to be used for applications such as online weighted model combination.

21 February 2020, 13:00-13:50 in K303 (Knowledge Centre, Engineering)

Febe de Wet, Thomas Niesler, Ewald van der Westhuizen, Astik Biswas (joint talk) - Automatic speech recognition and keyword spotting in under-resourced languages

Applications of automatic speech recognition technology have become prevalent, such as voice-enabled personal assistants, automatic dictation systems and hands-free in-car interfaces. However, this progress is limited to a small number of languages for which sufficient speech and text data are available for technology development. Most of the world’s languages do not fall into this category, however. In this talk we will present two projects focusing on resource and technology development for severely resource-constrained languages. The first involves the technical challenges of automatic speech recognition for code-switching among five South African languages, while the second deals with systems supporting humanitarian relief efforts in languages that are spoken in Uganda, Somalia and Mali.

6 March 2020, 13:00-13:50 in K303 (Knowledge Centre, Engineering)

Jan Buys

20 March 2020, 13:00-13:50 in K303 (Knowledge Centre, Engineering)

Noe Fouotsa Manfouo

24 April 2020, 13:00-13:50 in K303 (Knowledge Centre, Engineering)

Daniel B. le Roux

15 May 2020, 13:00-13:50 in K303 (Knowledge Centre, Engineering)

  • Rensu Theart
  • Kayode Olaleye

22 May 2020, 13:00-13:50 in K303 (Knowledge Centre, Engineering)

Charl van Heerden (SAIgen)