- Integrated ML/NLP with fMRI to study figurative-language processing in autistic and non-autistic groups.
- Aligned neural activity with BERT-derived semantic embeddings.
- Applied Ridge Regression for decoding (sentence from activity) and encoding (activity from embeddings).
- Applied ML classification methods to detect emotions from EEG signals in the SEED dataset.
- Boosted accuracy by 44.2% (SVM), 47.2% (KNN), and 29.9% (LR) via feature extraction and selection.
- Explored how Concatenate ReLU and Leaky ReLU activation functions mitigate plasticity loss in DQN.
- Prevented success-rate drops and neural-network neuron “death,” maintaining stable gradients.
- Sustained adaptability in non-stationary continual-learning tasks.
- Led a CRISP-DM data-mining pipeline to predict newborn birth weight and classify risk.
- Developed regression and classification models using ensemble, linear, tree-based, probabilistic, and neural methods.
