Research
Learning discriminative conditional models and inference on discriminative temporal chains (deriving discriminative propagation rules and learning multimodal conditional distributions based on Bayesian mixture of experts)
Variational Mixture Smoothing for Non-Linear Dynamical Systems (compact multiple hypothesis smoothing in a Bayesian framework)
Learning Continuous Generative Models using Non-Linear Manifold Embedding (continuous optimization over non-linearly learned manifolds)
Generalized Darting (fair but efficient sampling from an equilibrium distribution using its known minima)
Kinematic Jump Sampling (deterministic hypothesis generators based on interpretation trees and closed-form inverse kinematics)
Hyperdynamic Importance Sampling (focusing the samples near the saddles based on their gradient and curvature signature)
Eigenvector Tracking / Hypersurface Sweeping (deterministic trajectories linking minima via transition states)
Covariance Scaled Sampling (adaptively steering the sampling pattern based on distribution's shape/covariance)
Visual Modeling and Learning, Tracking, Pattern Recognition
Articulated Human Motion Estimation 3D model-based articulated motion extraction in monocular video sequences
Deformable Models incremendal adaptive model building, bundle adjustment, high-level grouping
Medical Imaging Classification and High-Dimensional Space Indexing pathology representation and classification
Computational Neuroscience models for learning and memory in the hippocampus and neocortex, neural networks
Topics I worked on earlier