Federated Learning

I'm currently working on Federated Learning(FL). In FL, each client trains its local model based on its own data, and then uploads the model to the server. Then the server aggregates the local models and broadcast the aggregated model to the clients. In this manner, the privacy is well kept as the data stays locally. Compared wit classic machine learning which brings data to codes, FL brings codes to data. But this brings up more problems to solve and interesting research topics. One of which is in the communication aspect.

Alpha-stable Distribution

Alpha-stable distribution is a kind of distribution to model the impulsiveness or heavy-taildness of the distribution. See my publications for more details.

Copula Theory

Copula is a powerful mathmatical tool that enables us to model the dependence seperatedly from modeling the marginals. See my publications for more details.