Thời gian: 14h Thứ 5, ngày 28/04/2022

Link online meet.google.com/nqs-ntdh-wnz

Abstract: Simulating neural networks has received a lot of attention in recent years, for example, there exist several grand projects such as the Human Brain Project in Europe, the Brain Mapping in Japan and the Brain Initiative in the United States. A complete simulated model will help reduce the cost of performing biological experiments and create an important premise before the actual experiments. In practice, there are many approaches to simulate neural networks. Here, we will focus on the point process approach, especially the simulation algorithm using the classical Ogata’s thinning method. However, it has been found to be too expensive for the huge network. In this talk, we will encounter this problem by presenting a new simulation method which combines the classical Ogata thinning method and a new type of Kalikow decomposition. We also focus to prove mathematically that new algorithms return a right intensity processes and present several numerical results.
This talk is based on the recent works with Patricia Reynaud-Bouret, Eva Locherbach, Alexandre Muzy and Paul Gresland