AI Pontryagin, or: How Neural Networks Learn to Control Dynamical Systems

The efficient control of complex dynamical systems has many applications in the natural and applied sciences. In most real-world control problems, both control energy and cost constraints play a significant role. Although such optimal control problems can be formulated within the framework of variational calculus, their solution for complex systems is often analytically and computationally intractable. In this talk, I will present AI Pontryagin, a versatile control framework based on neural ordinary differential equations that automatically learns control signals that steer high-dimensional dynamical systems towards a desired target state within a specified time interval. I will discuss examples that demonstrate the ability of AI Pontryagin to learn control signals that closely resemble those found by corresponding optimal control frameworks in terms of control energy and deviation from the desired target state. I will also discuss possible applications of AI Pontryagin in computational biology and medicine, where neural-network-based control frameworks can help solve a wide range of control and optimization problems, including those that are analytically intractable. To learn more see: Böttcher, Lucas, Nino Antulov-Fantulin, and Thomas Asikis. "AI Pontryagin or how artificial neural networks learn to control dynamical systems." Nature communications 13, no. 1 (2022): 333. https://www.nature.com/articles/s41467-021-27590-0 To view the slides in this video see: https://drive.google.com/file/d/1G8Gb6HKWoh4DN02aYyq1Dyp86CUxMfU3/ Patient Specific Cell-Cell Networks Suggest Important Links in Disease Progression David Gibbs Institute for System Biology, Seattle Cell-cell communication is involved with regulating inflammation, promoting proliferation and differentiation, tissue repair, and to guide cell migration in the body. Abnormal cell-cell communication can cause disease, and in the opposite direction, diseases can alter communication. Cancer, once seen as a disease of genetics, is now recognized as being inexorably connected to the complex host of cellular interactions within the tumor microenvironment, which shape tumor growth and response to therapeutics. One approach to studying cell interactions is through the use of quantitative network models. In this work, we combined multiple sources of data with a probabilistic method for computing patient level weighted networks that provide predictive features. In total, we constructed 9,234 weighted networks using the TCGA PanCancer data set, containing 64 cell types and 1,894 ligand-receptor pairs. Using robust statistics, informative network features can be found that are associated with disease progression. The entire collection of data, network weights, and results are stored in BigQuery database tables, hosted in the google cloud by ISB-CGC. To learn more see: Gibbs, David L., Boris Aguilar, Vésteinn Thorsson, Alexander V. Ratushny, and Ilya Shmulevich. "Patient-specific cell communication networks associate with disease progression in cancer." Frontiers in Genetics 12 (2021): 667382. https://www.frontiersin.org/articles/10.3389/fgene.2021.667382/full To view the slides in this video, see: https://drive.google.com/file/d/1bancegdpEc1SYBg6jo-9LfexCJ24AFe6/view?usp=sharing *Contents* 0:00 - Introduction: J Glazier 3:30 - Coming Up Next week 4:55 - Lucas Bottcher: AI Pontryagin, or: How Neural Networks Learn to Control Dynamical Systems 27:55 - David Gibbs: Patient Specific Cell-Cell Networks Suggest Important Links in Disease Progression 49:23 - Q&A Session Moderator: James A. Glazier, PhD, Indiana University, Bloomington If you found this video useful, please check out our other videos on computational modeling, infection and immunology: https://tinyurl.com/GLIMPRINTVideos Please consider joining our IMAG/MSM WG on Multiscale Modeling and Viral Pandemics: https://www.imagwiki.nibib.nih.gov/content/msm-viral-pandemics-meetings Please also consider joining the Global Alliance for Immune Prediction and Intervention: http://glimprint.org/