Diego Mesquita

Diego Mesquita

Since 2022, I am an assistant professor at FGV's EMAp. I am interested in Bayesian inference, deep learning, and distributed ML. The common thread here is to strive for greener (compute-efficient) AI. Lately, I am also working in applications to software engineering and blockchain analysis.

Follow me on twitter @wkly_infrmtive for somewhat regular updates and for occasional bad jokes. You can also contact me through diego.mesquita@fgv.br

Want to work with me? I'm always looking for excellent post-docs an PhD/MSc students.

Post-doc positions: send me an email with your CV, a brief motivation, and a pointer to your favorite work.
MSc/PhD students: In both cases, you need to sign up for EMAp's selective process. For PhDs, I strongly encourage you to contact me before applying.

I am committed to foster diversity and encourage folks from different backgrounds (e.g., nationality, sex, gender, and religion) to apply.

Selected publications

D. de Souza, A. Nikitin, S. T. John, M. Ross, M. A. Álvarez, M. P. Deisenroth, J. Gomes, D. Mesquita, C. Mattos Thin and deep Gaussian processes . NeurIPS 2023 [video]
T. Pereira, E. Nascimento, L. Resck, D. Mesquita, A. Souza. Distill n' Explain: explaining graph neural networks using simple surrogates . AISTATS 2023 [video]
A. Souza, D. Mesquita, S. Kaski, V. Garg. Provably expressive temporal graph networks. NeurIPS 2022 [video]
D. de Souza, D. Mesquita, S. Kaski, L. Acerbi. Parallel MCMC without embarrassing failures. AISTATS 2022 [video]
K. El Mekkaoui, D. Mesquita, P. Blomstedt, S. Kaski. Federated stochastic gradient Langevin dynamics. UAI 2021 [video]
D. Mesquita, A. Souza, S. Kaski. Rethinking pooling in graph neural networks.NeurIPS 2020 [video]
D. de Souza, D. Mesquita, J. Gomes, C. Mattos. Learning GPLVM with arbitrary kernels using the unscented transformation. AISTATS 2020 [video]
D. Mesquita, P. Blomstedt, S. Kaski. Parallel MCMC using deep invertible transformations UAI 2019

Preprints

T. da Silva, A. E. Silva, A. Ribeiro, A. Góis, D. Heider, S. Kaski, D. Mesquita Human-in-the-Loop Causal Discovery under Latent Confounding using Ancestral GFlowNets . under review [video]
P. Blomstedt, D. Mesquita, O. Rivasplata, J. Lintusaari, T. Sivula. J. Corander, S. Kaski Meta-analysis of Bayesian analyses. under review [video]

For a full list, have a look at my Google Scholar page.

Supervision

Post-docs

Eliezer Silva

PhD

Tamara Pereira

Master's

Alumni: Jamille Feitosa, João Duim, Eduardo Guimarães, Tuomas Aarnio, Tuomas Myllymäki, Antti Kumpumäki.