Machine learning research

The Romanian Institute of Science and Technology currently has about 25 scientists working on research in machine learning and artificial intelligence (AI).

Some current research topics are:

  • developing technologies for program synthesis or induction, using deep learning and other machine learning methods, with a focus on the automated development of web pages (HTML / CSS) or the synthesis of new types of autoencoders / compressors;
  • the design and analysis of novel training algorithms for neural networks in deep learning, by applying notions of Riemannian optimization and differential geometry.
  • studying how intelligent agents are able to learn about their environment by interacting with it, through curiosity-driven learning, using approaches such as deep learning, hierarchical reinforcement learning and meta-learning;

Previously, our scientists have developed several new learning rules for spiking neural networks. Components of such a learning rule have been implemented in Intel’s Loihi neuromorphic chip.

We welcome collaborations with companies where we can apply our know-how in deep learning and other machine learning methods. Contact us to discuss about this.

Our know-how in deep learning can be broadly applied in a large variety of fields. Some examples of topics we approached are:

  • automated generation of web pages;
  • detecting tumors in brain scans;
  • detecting anomalies in heart sounds;
  • defending against adversarial attacks in image classification, e.g. of traffic signs;
  • computer vision applications;
  • text analysis, e.g. sentiment detection.

Scientific publications in machine learning