Latest News

Research Interests

I am personally interested in the applications of machine learning models in medicine, particularly in addressing issues such as how medical data should be analyzed and how machines and humans can interact and complement each other.

In pursuit of the first interest, I have focused on developing deep generative models that can uncover the hidden correlations inherent in medical data. Specifically, my work has centered on leveraging intra-variable sequential correlations in temporal data and inter-variable correlations among different signals to gain valuable insights from both global and specific contexts.

Recently, I have been exploring novel approaches to the following questions related to my second area of interest: How can we integrate humans and machines into the decision-making process effectively? In a hybrid intelligence system that comprises both humans and machines, how can we determine when to rely on the machine and when to defer to human judgment? Moreover, how can we design a system that can learn from both humans and machines to improve decision-making?


*Co-first authors, order determined by coin flip.

Learning to Defer to Multiple Experts
Rajeev Verma*, Daniel Barrejón* and Eric Nalisnick
Artificial Intelligence and Statistics (AISTATS), 2023. [arXiv] [paper] [code] [poster] [bibtex]

On the Calibration of Learning to Defer to Multiple Experts
Rajeev Verma, Daniel Barrejón and Eric Nalisnick
In ICML 2022 Workshop on Human-Machine Collaboration and Teaming, 2022. [preprint] [bibtex]

Medical data wrangling with sequential variational autoencoders
Daniel Barrejón, Pablo M. Olmos and Antonio Artés-Rodríguez
In IEEE Journal of Biomedical and Health Informatics, 2021. [arXiv] [paper] [code] [bibtex] [poster] [slides]


  • Universidad Carlos III de Madrid
    Dept. of Signal Theory and Communications
    Room 4.3.B01
    Avenida de la Universidad, 30
    28911 Leganés, Spain
  • (+34) 916248839
  • Write me at !

Not work

I play guitar and piano. Music Lover.

Below some records I've been involved!

Design by HTML5 UP. Pictures are my own.