報(bào)告題目：EPR-Net for Constructiong the Landscape of Biological Systems

報(bào) 告 人：李鐵軍  教授 (北京大學(xué))

報(bào)告時(shí)間：2023年11月1日上午10：30-11：30

報(bào)告地點(diǎn)：數(shù)學(xué)科學(xué)學(xué)院A413    （騰訊會(huì)議 768-203-461）

內(nèi)容簡介：

We present the EPR-Net, a novel and effective deep learning approach that addresses a crucial challenge in biophysics: constructing potential landscapes for high-dimensional non-equilibrium steady-state (NESS) systems. The key idea of our approach is to utilize the fact that the negative potential gradient is the orthogonal projection of the driving force with respect to an inner product weighted by the steady-state distribution. Remarkably, the minimum of our proposed loss function coincides with the steady entropy production rate (EPR) formula in NESS theory. We also introduce an enhanced learning strategy for systems with small noise, and extend our unified framework to dimensionality reduction and state-dependent diffusion coefficients. The proposed approach is successfully applied to different biophysical examples.

報(bào)告人簡介：李鐵軍，北京大學(xué)數(shù)學(xué)學(xué)院教授，國家級(jí)稱號(hào)人才。研究領(lǐng)域?yàn)殡S機(jī)模型及算法，在復(fù)雜網(wǎng)絡(luò)、生物體系隨機(jī)動(dòng)力學(xué)、單細(xì)胞轉(zhuǎn)錄組數(shù)據(jù)分析領(lǐng)域做出了重要貢獻(xiàn)。

（撰稿：梁西銀  審核：張國）

                                              數(shù)學(xué)科學(xué)學(xué)院   

2023年10月23日