Belinda Tzen, Tengyuan Liang, Maxim Raginsky. Local Optimality and Generalization Guarantees for the Langevin Algorithm via Empirical Metastability. arXiv:1802.06439, accepted at COLT, 2018.
Tengyuan Liang, Weijie Su. Statistical Inference for the Population Landscape via Moment Adjusted Stochastic Gradients. arXiv:1712.07519, 2017.
Alexandre Belloni, Tengyuan Liang, Hariharan Narayanan, Alexander Rakhlin. Escaping the Local Minima via Simulated Annealing: Optimization of Approximately Convex Functions. Journal of Machine Learning Research W&CP, vol.40 (COLT), 240–265, 2015.