hiyenwong

Supervised Deep Multimodal Matrix Factorization (SD3MF) methodology for interpretable brain network analysis. Generalizes SNMTF from unsupervised single-graph clustering to supervised prediction over populations of multimodal graphs. Learns deep hierarchical factorizations with shared latent repres…

Distributed zeroth-order policy gradient for networked multi-agent reinforcement learning from human feedback (RLHF). Addresses scalability of preference-based RL to multi-agent systems without centralized training. Activation: distributed multi-agent RLHF, zeroth-order policy gradient, networked M…

MōLe-Λ methodology for learning coupled-cluster response states. Extends Molecular Orbital Learning (MōLe) to predict full CCSD response state by jointly learning T and Λ amplitudes from localized Hartree-Fock orbitals. Provides CC-quality energies, forces, dipoles, polarizabilities, electron densi…

Bayesian theory of attention pattern emergence in transformers — derives closed-form posterior over attention matrices, reveals first-order phase transitions in training data amount for copy head emergence, contrasts softmax vs linear attention behavior.