Abstract: In recent years, the scale and dimensionality of data associated with machine learning applications have seen unprecedented growth, spurring the BIG DATA Research and Development. Many machine learning methods are formulated as an optimization problem, including support vector machine, deep neural networks, etc. How to solve the optimization problem efficiently remains a challenging problem for large-scale high-dimensional data. Recent research and development revolve around designing fast stochastic optimization algorithms for solving big data machine learning problems that rely on smoothness and strong convexity assumptions of the problem. In this talk, I will talk about our recent research on developing faster stochastic optimization algorithms without imposing strong assumptions about the problem.