On the afternoon of December 15, 2020, Mahendra Mariadassou, Junior Researcher at Institute National de la Recherche Agronomique(France), was invited to give an online report entitled "The Poisson-lognormal Model for the Joint Analysis of Species Abundances". The report was hosted by Professor Liang Guo, deputy dean of the school. Teachers Zhang Yusen, Zhou Li, Wang Xiaoqiang and graduate students attended the report and participated in the online discussion actively.

Poisson-lognormal model is a kind of multivariate mixed generalized linear model based on Poisson distribution, which has been widely used in microbial ecology. In the talk, Pr. Mahendra Mariadassou introduced the structure, characteristics, and analysis methods of counting data from the perspective of datasets. Next, he introduced two kinds of PLN models and their inference methods based on probabilistic PCA and sparse network structure, respectively. He suggested that the potential interaction between species (competition, mutualism, parasitism, etc.) and the structuring factors (habitat type, nutrient availability, etc.) could be depicted by inferring the parameters in the model, which enables to model the joint abundance of all species from the same community. He also proposed that it is effective to handle   statistical and computational issues on large datasets using variational inference and convex optimization. At the end of the report, Pr.  Mahendra Mariadassou discussed with teachers and students on how to improve the model and inference of high-dimensional data. From the well-organized and fascinating talk, teachers and students benefited a lot. They also had a better understanding of the PLN model, which shall be a good guidance to the further research.