Hi, my name is Timi! My interests lie in the intersection of Generative AI and Data Science and I love building applications. For example, thinking of ways to build bespoke RAG frameworks, or finding interesting ways to use AI agents to analyse complex data. Academically, I completed my MSc. in Statistics, where I wrote my thesis on the adversarial robustness of Spiking Neural Networks, intersecting AI Safety and neural networks.
Research and Papers
Adversarial Robustness of Rate-Encoded Spiking Neural Networks
As part of my Master's thesis, I wrote a technical report studying the resistance to adversarial attacks of rate-encoded Spiking Neural Networks (SNNs) across various white-box and black-box attacks. This study explores the robustness of SNNs in different adversarial scenarios and proposes methods to enhance their security and reliability.
Modelling Heart Conditions and Train Delays Using Machine Learning Methods
Alastair Harrison, Timi Folaranmi, Ying Zhan, Weiyun Wu
The first study looks at the question of: To what extent can heart conditions be predicted from ECG readings? The second study looks at a machine learning approach to predicting train delays, and conducting all analysis in R.
An Exploration into Support Vector Machines (SVMs) with comparisons to other Classification Methods
Jake Dorman, Timi Folaranmi, Anas Almhmadi, Rishabh Agarwal
A technical report on the background, performance and evaluation of Support Vector Machines in solving general classification problems (in Python), compared with other classification methods.
Learning Projects
Neural Networks
Sequence classification with LSTM Recurrent Neural Network, using Keras
Explored movie review sentiment classification problem using LSTM Recurrent Neural Network, with Keras.
Classification using neural networks in PyTorch
Neural network for basic multi - class classification.
Computer Vision
Vision Transformer (ViT) from scratch in PyTorch
A paper implementation of the (original) Vision Transformer (ViT) architecture using PyTorch. Applies convolutional neural network (CNN) method. It follows from code by Daniel Bourke
Natural Language Processing
Text Summarisation, using NLP
A project exploring text summarisation, applying tokenisation and extraction method.
Recommendation Systems
Collaborative Filtering system
Implementation of a collaborative filtering (CF) based system looking at user - based and item based CF and Alternating Least Squares (ALS) on a restaurant problem.
Paper Implementations
I tend to learn better when trying to apply concepts from papers. Below are some basic paper implementations directly using labml and most of their implementation code: