Projects

Recent projects I have worked on.

Recovering Mesh Centroids from LiDAR Point Clouds using SE(3)-Equivariant Graph Networks

Recovering Mesh Centroids from LiDAR Point Clouds using SE(3)-Equivariant Graph Networks

Fagprojekt (DTU, 02466): building rotation- and translation-equivariant graph neural networks to estimate object centroids from sparse LiDAR point clouds, and testing whether built-in symmetry beats data augmentation.

Geometric Deep Learning
Graph Neural Networks
Equivariance
PyTorch
Point Clouds

Examination of Prompt-Induced Biases in Moral Evaluations by GPT-4o

Examination of Prompt-Induced Biases in Moral Evaluations by GPT-4o

Project in Statistical Evaluation (DTU, 02445): testing whether the sentiment and perspective of a prompt sway GPT-4o's moral judgments of interpersonal conflicts, using a designed experiment and ART ANOVA.

Large Language Models
Bias
Statistical Evaluation
ANOVA
Python

Can Heart Rate Predict Frustration? Evaluating Predictive Models

Can Heart Rate Predict Frustration? Evaluating Predictive Models

Individual assignment (DTU, 02445): a group-aware model comparison testing whether heart-rate summary statistics can predict self-reported frustration, and finding, honestly, that they cannot.

Machine Learning
Statistical Evaluation
Cross-Validation
Model Comparison
Python

iBeat: CNN Beat Detection in Short Music Snippets

iBeat: CNN Beat Detection in Short Music Snippets

First-semester 3-week project (DTU, 02461): my first hands-on work with CNNs, comparing waveform- and spectrogram-based convolutional networks for detecting beats in 10-second music clips.

Convolutional Neural Networks
Audio
Signal Processing
PyTorch