CITIC

Precise Statistical Estimates with Insufficient Samples and Reinforcement Learning at CITIC's TIC Talk Breakfasts

30/04/2024 - CITIC

Researchers Naomi Diz Rosales and Alejandro Fernández Camello headlined a new session of CITIC’s TIC Talk Breakfasts this Tuesday, April 30, focusing on AI and statistics. 

“Socioeconomic and Biomedical Challenges in Small Areas” was the title of Naomi Diz’s presentation, in which she addressed obtaining precise estimates of indicators of interest when the sample sizes in the target areas are insufficient. The challenge of “Small Area Estimation” is crucial for addressing issues such as the proportion of poverty by province and gender or the occupancy of intensive care units in pandemic contexts by healthcare area and day. 

The complex socioeconomic and healthcare context experienced in recent years has heightened the need for institutions to have more detailed information for monitoring inequalities, implementing measures, or distributing funds. Hence the importance of Naomi Diz’s research on small area estimation. 

Researcher Alejandro Fernández Camello presented an “Introduction to Reinforcement Learning,” examining its principles and how it is developed in different environments, distinguishing between reinforcement learning and machine learning. He also discussed specific examples such as “Voyager,” an agent that uses ChatGPT to learn to play Minecraft, and AutoGPT, designed to automate tasks on computers using ChatGPT. 

The TIC Talk Breakfasts are held monthly and aim to enhance cross-disciplinary collaboration and promote potential synergies among research staff.