Read on Twitter
2nd paper of the day with friends from @GRAPPAinstitute, incl. @C_Weniger, @NilanjanBanik and @gfbertone! We haven't solved protein folding, but I am quite excited about this study in which we show the full potential of simulation-based inference for grand scientific problems
https://twitter.com/gfbertone/status/1333726886543945728
https://twitter.com/gfbertone/status/1333726886543945728
In this work led by my student @joeri_hermans, we use amortized approximate ratio estimators ( http://proceedings.mlr.press/v119/hermans20a/hermans20a.pdf) for the statistical analysis of the observed perturbations in the density of stellar streams. (Image credits: C. Bickel/Science)
Long story short: clumps of dark matter should perturb stellar streams. Therefore, their analysis can indirectly set stringent constraints on the mass of the dark matter particle. (Image credits: D. Erkal, see https://people.ast.cam.ac.uk/~derkal/files/pal5_pr/ for other cool animations!)
Or said otherwise, the analysis of stellar streams could help decide whether dark matter is either cold or warm (CDM vs WDM), a long-standing question in cosmology! (Image credits: Bonaca et al, 2014)
In our study, we show that recent progress in simulation-based inference algorithms, in which sufficient summary statistics are automatically machine learned from simulated data, could provide elements of answers by setting stronger constraints on the WDM mass.
In particular, and under simplifying assumptions of the simulation model (*big disclaimer here*), our preliminary results for GD-1 suggest a preference for CDM over WDM.
From a methodological point of view, the paper also includes a suite of diagnostics we designed to assess the correctness of the inference results -- something currently quite underappreciated in the today's SBI litterature.
Finally, our experiments and plots are all directly reproducible on @mybinderteam at https://github.com/JoeriHermans/constraining-dark-matter-with-stellar-streams-and-ml. Enjoy!
