This morning, the twentieth TIC Talk Breakfast was held, a monthly event organized by CITIC. On this occasion, the presentations were given by researchers Jorge Castiñeiras Rella and Daniel Iglesias Morís. 

Jorge Castiñeiras’s work, titled “Methodologies for addressing credit risk,” starts from the premise that credit risk analysis depends on customer payment behavior and changes according to the economic cycle and market trends, while also being cost-sensitive because the results depend on the loan amount. Additionally, there is only information available on approved transactions, which can lead to unfair biases and missed opportunities. In this regard, he proposes new dynamic learning strategies that enhance online learning and bandit algorithms, making them more cost-sensitive. 

Daniel Iglesias Morís presented his work “Improving lung disease screening with generative models of medical imaging.” In the talk, he discussed various data augmentation techniques that generate synthetic images applied to 2D chest radiography, using different generative model architectures such as GANs and latent diffusion models. This approach enhances the effectiveness of automatic methods for detecting lung diseases.