报告题目：Non-i.i.d. Statistical Learning Theory
报告人： 张超 博士 （大连理工大学）
In this talk, we first review some classical results of statistical learning theory, which are almost built under the sample-i.i.d. assumption. However, this assumption will not be naturally satisfied in practice. Thus, it is necessary to study the properties of the machine learning problems for non-i.i.d. settings. Some typical non-i.i.d. learning problems will be presented in this talk, for example, domain adaptation and ERM principle for stochastic-process samples. The main theoretical results are given in term of the following aspects: the generalization bounds, the consistency and rate of convergency, etc.