Click a card to open the repository on GitHub. The list below highlights recent work and experiments.
Graphormer DGL
Sign-invariant positional encoded Graphormer with spatial/centrality/edge encoders; training scripts and ZINC dataset utilities.
GIN DGL
Graph Isomorphism Network (GIN) with DGL/PyTorch; trains on MUTAG, PTC, NCI1, PROTEINS, COLLAB, IMDBBINARY, IMDBMULTI with accuracy plots.
RGCN DGL
Relational GCN for entity classification (DGL/PyTorch); supports AIFB, MUTAG, BGS, AM with best-validation accuracy plots.
Graph-Deeplearning-GAT DGL
GAT experiments and attention-based graph models.
RUM-Graph-nets
Non-convolutional GNN via random walks + RNN (PyTorch Geometric); addresses over-smoothing/over-squashing with scalable memory-efficient design.
GCN-PYG
GCN with PyTorch Geometric; runs on Cora/Citeseer/PubMed with best test accuracy across seeds plotted.
interpretable-GNNs DGL
Implementation of Graph Neural Additive Networks (GNAN) with DGL; includes feature-dependent distance transform and notebook demos.
multi-arm-bandits
ε-greedy multi-armed bandits with JAX; plots average reward and optimal-arm selection over steps.
on-polciy-mote-carlo-with-jax
On-policy Monte Carlo control in JAX (Blackjack env); includes Plotly heatmap of optimal actions and notebook walkthrough.
value_n_policy_learning
Soft Actor-Critic (SAC) with JAX/Brax/Equinox; uv-managed deps, training loop, and interactive Plotly returns in `plots/`.