Seminar "Intersubjective agreements"

Towards the Plurinational Bayesian Congress 2023

The webinar is an opportunity to discuss a wide range of topics. We focus on Bayesian aspects of practical, theoretical and philosophical applications in all branches of science and in all productive and social activities. If you want to participate, send your proposal through the form: https://bit.ly/SeminarioComunidadBayesiana.

Seminar 4.

Bayesian Learning with Wasserstein Barycenters

• • Gonzalo Ríos. Chief Scientific Officer at NoiseGrasp Spa.
• • When: During September (to be announced)
• • Where: Virtual (to be announced)
• • Abstract: It will be presented a model selection paradigm based on the optimal transport criterion: the barycenter under the Wasserstein distance. The selected model will be characterized and some tools for its calculation will be provided.
• • Bio: Head of the Data Science area for the development of predictive marketing models at NoiseGrasp Spa. Trained at the University of Chile as Doctor in Engineering Sciences, Master in Computer Science, Bachelor in Mathematical Sciences, Civil Engineer in Mathematics. For a decade he was a professor at the University of Chile in the departments of mathematics, computer science and engineering.

Seminario 3.

Estimation of cosmological parameters through Bayesian Statistics.

• • Francisco Xavier Linares Cedeno. Posdoctoral researcher at Instituto de Física y Matemáticas - Universidad Michoacana de San Nicolás de Hidalgo, México.
• • When: June 1 (Thursday) at 4pm GMT-3 (Argentine time)
• • Where: https://www.youtube.com/@bayesdelsur
• • Material: Science popularization text Cosmología e Inferencia Bayesiana.
• • Abstract: This talk explores the process of inferring cosmological parameters using the Bayesian Theorem. Firstly, it introduces the Einstein equations as the basis for constructing the standard cosmological model, known as LambdaCDM. To analyze Supernova observation data, it defines the distance calculation for luminous objects, referred to as the theoretical distance modulus. Secondly, it presents the information obtained from supernovas and the procedure for parameterizing the observable distance modulus based on their intrinsic properties. Thirdly, using the Bayesian Theorem, it establishes a statistical comparison between the predictions of the LambdaCDM model and the distance modulus derived from supernovas. Specifically, it imposes constraints on the total matter parameter of the universe and poses the question: "What is the most probable value for the amount of matter in the universe, given the supernova data?" Finally, it discusses the application of Bayesian statistics in other cosmological contexts.
• • Bio: Francisco holds a licentiate degree in Physics from Simón Bolívar University (USB) in Venezuela. He obtained his Ph.D. in Physics from the University of Guanajuato in Mexico. He has conducted research stays at various Mexican universities, including ICF-UNAM, IFM-UMSNH and DI-UG. Currently, he is a postdoctoral researcher at the Institute of Physics and Mathematics at the University Michoacana de San Nicolás de Hidalgo (IFM-UMSNH) and a member of the Legacy Survey of Space and Time (LSST-México) at the Vera Rubin Observatory. Francisco specializes in analyzing the dynamics of the cosmos using a Bayesian approach to probability.

Seminar 2.

Impact of Abdala SARS-CoV-2 vaccine against SARS-CoV-2 on the incidence, severity and mortality due to COVID-19 in Havana, Cuba.

• • Kenia Almenares. Escuela Nacional de Salud Pública de Cuba.
• • Language: Spanish talk
• • When: May 11 (Thursday) at 4pm GMT-3 (Argentine time)
• • Where: Live at https://www.youtube.com/@bayesdelsur
• • Co-authors: Pedro Más Bermejo DrCs., Lizet Sánchez Valdés DrC., María J. Vidal Ledo DrC.
• • Material: Article Impacto y efectividad de la vacuna Abdala en la provincia Matanzas ante la enfermedad sintomática y la muerte por COVID-19
• • Abstract: The Cuban vaccine Abdala was designed by the "Centro de Ingeniería Genética y Biotecnología de Cuba", to control the COVID-19 epidemic. It was implemented in a vaccination campaign in persons ≥ 19 years old, in Havana, during May-July 2021. Objective: To evaluate the impact of vaccination on the incidence, severity and mortality due to COVID-19, in Havana, Cuba, from July to October 2021. Methods: A longitudinal causal impact study was conducted in Havana, Cuba. The entire Havana population (2,137,936 inhabitants) was included and the Bayesian Structural Time Series (BSTS) model was applied to obtain the impact of vaccination with the Abdala vaccine in the short term, with the calculation of the difference (counterfactual) between the observed time series of incidence, severity and mortality, and what would have happened if vaccination had not been administered. The time series of incidence, severity and mortality of Camagüey and Holguín provinces in the same period of time were used as control series (probability intervals of 95.0% were used). Results: In the absence of the intervention, an average weekly incidence value of 151 cases was expected, instead, 22 cases were observed, with a relative reduction of 54.0% (95.0% of the counterfactual interval was -70;-38). Severity did not show significant changes in the short term. Mortality, however, showed an expected value of about 3 deaths per week on average, the observed value was 2, with a reduction from the time of intervention of 36.0%, interval (-56;-14). Conclusions: The impact of vaccination evidenced a reduction in the indicators of incidence, severity and death due to COVID-19 in Havana, Cuba.

Seminar 1.

Modelling the dynamics of the primary visual cortex through recurrent neural network trained for probabilistic inference through sampling

• • Rodrigo Echeveste. CONICET researcher, from sinc(i) institute, Santa Fe, Argentina
• • Language: Spanish talk
• • When: May 4 (Thursday) at 4pm GMT-3 (Argentine time)
• • Where: Live at https://www.youtube.com/@bayesdelsur
• • Material: Article Cortical-like dynamics in recurrent circuits optimized for sampling-based probabilistic inference
• • Abstract: In recent years, Deep Learning has gained position as a tool for explaining the behavior of natural neural systems. However, these models lack dynamic behavior and often do not properly represent the uncertainty of their predictions. Our cerebral cortex presents rich dynamics, allowing to navigate a world in which the information we receive from our senses is always partial and incomplete. In this talk, I will be discussing how to train recurrent neural networks (RNN) under biological constraints. This allows approximating an ideal bayesian observer. Surprisingly, typical dynamic features of the visual cortex, such as gamma oscillations and transient responses with overshoot, emerge from this purely computational task.These types of models allow us to advance our understanding of the links between physiology and sensory perception.
• • Bio: Physicist from the Instituto Balseiro, with a PhD in Natural Sciences (Goethe University, Germany) collaborator at the Computational and Biological Learning (CBL) group at the Cambridge University (UK), CONICET research in one of the best AI Institutes in Latin America, sinc(i) Santa Fe, Argentina (https://sinc.unl.edu.ar/ ). He is a professor at the Universidad Nacional del Litoral (UNL) and coordinator of the Commission for the Federalization of the Argentine Society of Neuroscience Research.