報告題目：Statistical Learning of Large-scale, Sparse and Multi-source data
報 告 人：操龍兵 教授
Real-life big data is often very large, sparse, and multi-sourced. Such big data challenges widely appear in almost all big data applications including finance, communications, e-commerce, and social networks. Traditional statistical learning methods face significant problems because of the intensive mathematical computation and inference required. A recent direction in statistical learning is to develop efficient statistical methods to learn complex big data, requiring handling the bigness, high sparsity, heterogeneity and coupling in both observable and hidden spaces. Accordingly, this talk introduces some new statistical models for tackling such challenges on large, sparse and multi-source data with efficient Bayesian inference methods, in addition to its applications to large-scale and sparse recommendation and some new directions of statistically learning complex big data problems.