Hamilton-Jacobi Equations for Statistical Inference Problems
Time: 2020-09-28
Published By: He Liu
Speaker(s): Hong-Bin Chen (Courant, NYU)
Time: October 14 - October 15, 2020
Venue: Room 77201, Jingchunyuan 78, BICMR
Time: October 14 (Wednesday), 13:15--14:45
October 15 (Thursday), 15:15 -- 16:45
Note: There is also a probability seminar talk by Hong-Bin Chen on the same topic on Monday October 12, 2--3pm, in School of Mathematical Sciences. The seminar will be an overview of the problem and recent developments. More information of the seminar can be found here: http://bicmr.pku.edu.cn/~xinyili/probseminar.htm.
October 15 (Thursday), 15:15 -- 16:45
The aim is to give an overview of the Hamilton-Jacobi equation approach in the study of the limit of free energy in mean-field disordered systems. This approach will be demonstrated in relatively simple models from statistical inference. The free energy associated with an enriched Hamiltonian will be shown to converge to a unique solution of a certain Hamilton-Jacobi equation.
In the first lecture, we introduce the setting and basic tools. Then, we will show the free energy satisfies an approximate Hamilton-Jacobi equation. Notions of solutions will be discussed. In the second lecture, well-posedness of the equation will be studied, and the proof of the convergence will be sketched.Note: There is also a probability seminar talk by Hong-Bin Chen on the same topic on Monday October 12, 2--3pm, in School of Mathematical Sciences. The seminar will be an overview of the problem and recent developments. More information of the seminar can be found here: http://bicmr.pku.edu.cn/~xinyili/probseminar.htm.