Full Professor, Massachusetts Institute of Technology
2 papers at NeurIPS 2025
A new lower-bound for representation learning connecting Jensen–Shannon and Kullback–Leibler divergences.
We present PolyPose, a simple locally-linear method for deformably aligning a high-resolution preoperative 3D volume to as few as two intraoperative 2D X-rays without explicit regularizers or hyperparameter optimization.
