Research

Research is the ability to dive deep into complicated topics & derive unique insight 

Here are some research projects I have undertaken

Principia Musica
A mathematical decomposition of music. Analogies are often drawn between music and mathematics, alluding to the underlying structure of music.
 
It then naturally follows that if we consider music as a high-dimensional numerical representation, we ought to be able to apply statistical/mathematical methods, perform inference and model music.
AEVB for LDA
A challenge of modern data science is the diversity of the data: data types like text & images often translate to large, sparse numerical representations - which are often impractical for computation using traditional statistical techniques
My honour thesis we explored the viability of scaling Variational Bayes (VB) with a deep autoencoder (AEVB) to learn latent topic distributions (LDA)
We published the paper at the FAIR (Forum for Artificial Intelligence Research) conference & I presented my work in December 2019
Evolutionary Computation
An Implementation of Metaheuristics: Genetic Algorithms, Simulated Annealing & Particle Swarm Optimization to find approximate solutions to a Knapsack problem.
 
These intelligent optimization algorithms effectively search any high-dimensional domain to find good approximate global maxima/minima.  Though applied to knapsack (a discrete problem set) all techniques are easily modified to handle continuous search domains. In fact, these are applicable to any problem that can be posed mathematically.
 
Underlying philosophy: Given a sufficiently complicated problem set, write an algorithm at writes an algorithm.
BAMM
This paper introduces BAMM: a Bayesian Adaptive Mathematical framework for Malaria transmission dynamics.
 
BAMM begins by utilizing an existing malaria mathematical compartmental model (Yang [6]) & informs the parameterization of the model in a Bayesian setting by deriving an indirect mapping between both social-economic factors & medical data and the relevant transmission dynamics posterior parameter estimates. The framework provides a software tool in the form of a dashboard that allows researchers to posit & simulate any number of feasible scenarios. Making assessing either isolated events or the interaction effect of events trivial.
 
The framework allows for hypothesis testing on a wide variety of topics including (but not limited to): temperature changes, access to treatment, changes in inequality measures & growth in GDP.

Call

T: 084 496 8779

Follow me

  • Instagram
  • Facebook Social Icon