Nominations for 2022 Stan Governing Body

Stan Governance
13

Dear Community,

I am Caetano Souto Maior, and I would like to nominate myself for a two-year term in the Stan Governing Body.

I am currently a Postdoctoral Fellow in the Applied Statistics group at the Basque Center for Applied Mathematics (BCAM), where I am using Hamiltonian Monte Carlo (and improvements on its basic formulation) with nonlinear dynamic models.
My background, nevertheless, is in the natural sciences, where I have been not only defending but also applying bayesian inference for almost a decade.
I am faculty at the FAES Graduate School @ NIH teaching Linear Algebra with Applications in Statistics; Single; and Multivariate Calculus, as well as a certified trainer at The Carpentries. I have also volunteered as a translator for Raspberry Pi.

In that context, my intended contribution to the Stan Community is threefold:

  1. Help HMC algorithms to become more broadly understandable: Make education-oriented implementations of HMC, particularly with functional programming style that more closely encode the essential mathematical/statistical functions/descriptions of bayesian inference and MCMC.
    The expected result of this goal is to have simpler, more concise algorithms that allow the inner workings of MCMC to be easily explained to both newcomers and users of a broad range of backgrounds – also, the goal is to make more advanced features (like those implemented in Stan) to be less opaque;

  2. Expand the use of HMC in the natural sciences: Contribute to the dissemination of Hamiltonian Monte Carlo in process-oriented (often dynamic, nonlinear) modeling, with all of its constraints and specificities.
    The expected result is that researchers (biologists and life scientists especially) will become boradly aware of the power of bayesian inference and HMC; conversely, statisticians should be able to recognize the associated difficulties;

  3. Benchmarking: Build on goals 1 and 2 to implement different samplers applied to classical mechanistic models (like infectious disease dynamics with SIR-like models) as case studies oriented towards teaching/training.
    The expected result is a sort of benchmark for general sampler performance in more realistic models (e.g. Gradient-based vs. Metropolis-Hastings schemes), as well as for the computational cost of some features that make Stan user-friendly (e.g. auto-differentiation vs. analytical/numerical gradients, static vs. dynamic step numbers or intermediate approaches).

I also second proposals of some of the other nominees, especially the increased representation of the Global South – as a native Portuguese speaker living in a Spanish-speaking country I would be happy to contribute to efforts to translate any material to the two languages, spoken by around 800 million people worldwide. I am also eager to work with other members with different backgrounds and views, both to define priorities and improve efficiency of efforts, as well as to learn about other aspects of governing a successful project like Stan.

I am available to answer and questions or concerns, as well as to elaborate on the proposals summarized in points 1-3.

best,
Caetano

10 Likes