Mainly interested in innovative applications to deep learning, especially vision related ones and expanding unsupervised learning via research in Variational Autoencoders and Generative Adversarial networks. Also, very interested in the area of explorative reinforcement learning, which can provide general purpose models that can then be applied to more specific areas.
Theoretical notions behind computer science. In particular, automated reasoning and its relation to ontological management of data. Additionally, the mathematical and theoretical study of software modularity and other more general notions of modularity and elasticity.
The exploration of learning as both a tool to create solutions for data-rich problems and for the understandings of the underlying mechanisms that make learning, in a machine context, possible and plausible. This line of thought currently manifests as model analysis and manipulation for either robust representation learning or transfer learning.
Representation learning using deep neural networks. Representing sets, with applications to information retrieval, content-based recommendation and generative models. Adversarial learning for fair decision making.
Machine learning and computational semantics to study language, including language-independent representations of meaning, formal grammars and sentiment analysis. Working on mapping syntax constituents across different languages into a joint embedding space. Also looking at morphological variations and cultural marks, especially different ways of expressing happiness, sadness, humour or irony.
Bayesian approaches to deep learning, approximate inference, probabilistic programming languages and the application of these techniques to real world problems requiring the understanding of uncertainty.
Computational linguistics, natural language processing, cognitive modelling, complex networks. Currently focused on linguistic alignment & spreading phenomena in social media, & agent-based models of language change.