Currently I hold research in three main areas:

  1. Multi-objective optimization: I’m interested in how to make the optimization simpler, use surrogates to estimate the objectives and reduce the number of function evaluations, and how to learn dynamics of objectives that change through time. I’m also interested in innovative uses of multi-objective optimization, like replacing the mean loss with a hypervolume, and using the Pareto solutions to build an automated decision maker.
  2. Neural networks: In this area, my research is mainly focused on unsupervised learning and deep architectures, since they pose a more difficult but flexible problem. The main issues I try to solve are how to schedule the training in order to obtain better local minima, how to incorporate new samples without having to retrain the whole model, how to create high-level representations of data, and how to perform predictions using these high-level representations.
  3. Probabilistic models: Statistics is a major tool and inspiration for my research, and my focus in on generative models, which try to explain how the data is created, nonparametric models, which expand their capacity based on the data provided and avoids training issues, and online models, which deal with increasing data sets.

Although this isn’t an extensive list of my interests, it describes in a high-level in which problems I spend most of my time. For instance, I’m still very interested in robotics, which I started researching as an undergraduate, but my research in this field is on hold.