class: title background-image: url("figure/cover-transparent.png") background-size: cover <div id="title"> Beyond Black Box Simulation </div> <div id="subtitle"> ASA SSC Mini-Symposium <br/> [4 | November | 2023] </div> <div id="links"> Paper: https://go.wisc.edu/833zs8 Code: https://go.wisc.edu/7222i9 Slides: https://go.wisc.edu/fg7wr4 </div> <img src="figure/logo.svg" width=50 id="logo"/> <div id="bio"> <strong>Kris Sankaran</strong>, UW-Madison<br/> joint work with, <br/> <strong>Susan Holmes</strong> Stanford University </div> --- ### New Lingua Franca of Science .pull-left[ 1. Simulators have emerged as a general problem-solving device across various domains, many of which now have rich, open-source libraries. 2. Where is the interface with statistics? - Experimental design, model building, and decision-making. ] .pull-right[ .center[ <img src="figure/esm.png"/> ] The E3SM is used for long-term climate projections. ] --- ### New Lingua Franca of Science .pull-left[ 1. Simulators have emerged as a general problem-solving device across various domains, many of which now have rich, open-source libraries. 2. Where is the interface with statistics? - Experimental design, model building, and decision-making. ] .pull-right[ .center[ <img src="figure/splatter.png"/> ] Splatter generates synthetic single-cell genomics data. ] --- ### Grammar of Generative Models Transparent simulators can be built by interactively composing simple modules. Probabilistic programming has simplified the process. .pull-three-left[ <img src="figure/modules.jpeg" width=700/> ] .pull-three-right[ a. Regression <br/> b: Hierarchy <br/> c: Latent Structure <br/> d: Temporal Variation ] --- ### Discrepancy and Iterability By learning a discriminator to contrast real vs. simulated data, we can systematically improve the assumed generative mechanism. .center[ <img src="figure/iterability.jpeg" width=730/> ] --- ### Experimental Design Renaissance Let's consider a microbiomics case study: To block or not to block? * Blocking removes person-level effects... * ...but increases participant burden. <img src="figure/blocking_simplex.png"/> --- ### Simulation to the Rescue How can we navigate trade-offs like this? Simulate! .center[ <img src="figure/blocking.jpeg" width=840/> ] Simulators provide data for more precise decision-making. --- ### Covasim Following the outbreak of COVID-19, the research community came together to build simulators that could inform pandemic response. * E.g., "What would happen if we held classes remotely for two weeks?" .center[ <img src="figure/covasim.png" width=700/> ] --- ### Covasim Covasim is an example of an agent-based model. Starting from local interaction rules, it lets us draw global inferences. <img src="figure/emulation.jpeg"/> Statistical emulators mimic the relationship between input hyperparameters and output data, substantially reducing the computational burden. --- ### Conclusion We have many more examples, like the evolution of mimicry in butterflies, longitudinal study design, the duality between agents and particles, … * Paper Link: [https://go.wisc.edu/833zs8](https://go.wisc.edu/833zs8) * Code (R + Python + NetLogo): [https://go.wisc.edu/7222i9](https://go.wisc.edu/7222i9) **Inference and imagination**: Statistical calibration grounds us in reality while generative tinkering encourages us to imagine. --- ### Acknowledgements * Members of my research group - Hanying Jiang, Shuchen Yan, Zhuoyan Xu, Kaiyan Ma, Margaret Thairu, and Mason Garza * Funding mechanism: NIGMS R01GM152744