為推動我校人工智能學(xué)科建設(shè)，更好地讓學(xué)校相關(guān)老師更加深刻理解，學(xué)習(xí)人工智能技術(shù)和學(xué)術(shù)交流，軟件工程學(xué)院有幸邀請到法國雷恩第一大學(xué)Jean-Louis Coatrieux教授（國家千人專家），Régine Le Bouquin Jeannès教授和東南大學(xué)計(jì)算機(jī)學(xué)院副院長舒華忠教授來我院進(jìn)行訪問，學(xué)術(shù)交流和研討，歡迎對人工智能感興趣的老師和同學(xué)們踴躍參加。
報(bào)告時間：2019年4月18日（周四）9:10-11:10
報(bào)告地點(diǎn)：科技樓2號樓423會議室
報(bào)告簡介：
報(bào)告人：Jean-Louis Coatrieux教授
報(bào)告題目：What beyond Deep Learning？(人工智能通往何方?)
Abstract: Major research programs have been launched in recent years on what is called Artificial Intelligence (AI) all over the world based on the key advances brought by Deep Learning (DL) methods. This lecture will consist in two parts. First, we will look at the status, expectations, and perspectives offered by DL with some emphasis put on computer vision, speech and language translation. Their limitations will be also discussed regarding data, algorithms and other major issues. The current trends and some hot topics will be examined as well. Second, some paths will be sketched concerning the future. They aimed at answering the questions: how to go from DL to a new generation of systems capable to abstract, reason, explain, all features being essential components to capture knowledge and to approach the true “Artificial Intelligence” challenge?
報(bào)告人：Régine Le Bouquin Jeannès教授
報(bào)告題目：Dynamic causal modelling to infer changes in brain connectivity (動態(tài)因果模型推斷大腦聯(lián)通性的變化)。
Abstract: During the last years, we have been developing our research on drug-resistant epilepsies, trying to identify the epileptogenic zone (EZ) defined as the area of cortex indispensable for the generation of clinical seizures. This EZ must be resected such that seizures are consequently suppressed, or at least attenuated under the constraint that post-surgical deficits are limited. Identifying the EZ and the distributed sites for a better understanding of the organization of the seizure in terms of origin and propagation mainly refers to effective connectivity (i.e. causal effects of one neural system over another one in neuroscience). In this context, our goal consists in inferring information flow among different brain structures by detecting/analyzing relations inside an ensemble of signals recorded on multiple channels. In general, approaches quantifying effective connectivity can be divided into two categories, model-free approaches and model-based approaches. Contrary to the research we have conducted in the past on model-free approaches such as Granger-Wiener causality and transfer entropy, during the recent years, we have been mainly focusing on model-based approaches which assume some priors about how intracerebral electroencephalographic (iEEG) signals are generated. Among these model-based techniques, Dynamic causal modelling (DCM), and more particularly spectral DCM, is a conventional one. The objective of DCM is to identify the optimal model structure, which is determined as the one with the maximum free energy in a set of predefined plausible model structures as we will see in this presentation.
