Research Code— Delay Embedding & GPR
1. Logistic Map (Methodology Test)
We first checked whether the method works on a simple chaotic system.
I made a short time series at a fixed parameter, rebuilt the shape with delay embedding, and trained a small GPR model to predict one–two steps ahead.
Goal: with very little data, can we recover the curve and its short-horizon behavior?
2. Hénon Map (Chaotic Dynamics)
Next, I moved to a harder 2D chaotic system.
I created 1D observations with different measurement choices, varied the embedding dimension and delay, trained GPR, and compared predictions.
Goal: see how the observation choice and data size affect stable reconstruction and forecasts.
3. Ecological Application (Salvinia–Weevil, Kakadu)
Finally, I applied the method to real ecological biomass data.
I cleaned and resampled the series, focused on the most complete sites (Jabiluka and Island), split into train/validation, and tested several embedding settings.
Goal: check how well the forecasts track rise/fall patterns and the overall trajectory shape in real world data.
Mentored Research
Reconstructing Chaotic Dynamics with Delay Embedding and Gaussian Process Regression and Applying the framework to ecological biomass data
Independent Research
Exploring Genetic Disparities and Predictive Modeling in Gastric Cancer: A Statistical and Machine Learning Approach Using TCGA Data
