Romanian Institute of Science and Technology

Artificial intelligence & machine learning research

Artificial intelligence & machine learning are a strategic focus of our institute. We are looking forward to collaborations with companies on applied machine learning.

The Loihi neuromorphic chip launched by Intel implements learning rules developed at the Romanian Institute of Science and Technology

The Loihi chip implements in hardware “synaptic eligibility traces”, components of a learning rule developed by RIST’s Răzvan Florian, which allow the implementation of this learning rule on the chip.

Transylvanian Machine Learning Summer School

Our institute supported, as partner, the first edition of the Transylvanian Machine Learning Summer School (TMLSS), which took place in Cluj, Transylvania, Romania in July 16-22, 2018. The focus of school was on Deep Learning and Reinforcement Learning. The school was organized by machine learning researchers from DeepMind (Răzvan Pașcanu, Viorica Pătrăucean), Romanian Institute of Science and Technology (Luigi Malagò, Răzvan Florian), and McGill University & DeepMind (Doina Precup).

Understanding the mechanisms of alcohol addiction by analyzing brain networks

Our scientists have worked on mathematical and network theoretical models that contribute to preventing relapse in alcohol addiction.

Nobel prize winner James Watson visited our institute and lectured at the Transylvanian Experimental Neuroscience Summer School

James Watson has been awarded the 1962 Nobel Prize in Physiology or Medicine for his contribution to the discovery of the molecular structure of nucleic acids and its significance for information transfer in living material.

A new class of spike train metrics

Cătălin Rusu and Răzvan Florian have developed a series of several new metrics for quantifying the differences between spike trains. The new metrics are inspired by the Pompeiu-Hausdorff distance.

Neuronal membrane resonance enables stable and robust gamma oscillations in the brain

Vasile V. Moca, Raul C. Mureșan and colleagues have recently shown that membrane resonance, a property of inhibitory neurons, may hold the key to the emergence of robust and stable gamma oscillations in the brain.

Intricate connections between the Julia and Mandelbrot sets

RIST’s scientist Marius F. Danca and his colleagues have discovered that the Mandelbrot set is not only the set of complex plane points for which Julia sets are connected, but also the set of all parameter values for which alternated Julia sets are disconnected.

The chronotron – a neuron that learns to fire temporally precise spike patterns

Răzvan Florian has shown how neurons can learn to process and memorize information that is entirely temporally coded, both as input and as output, in the precise timings of spikes.

Finding attractors of continuous-time systems by parameter switching

RIST’s Marius F. Danca has found that if the control parameter p, of a continuous-time nonlinear system belonging to a large class of systems, is switched within a set of chosen values in a deterministic or even random manner, while the underlying model is numerically integrated, the obtained attractor is a numerical approximation of one of the existing attractors of the considered system.

The brain’s timescale correlates to the timescale of visual stimuli

Our scientists have found that the internal timescale of the brain, i.e., the time window needed by neurons to encode a given aspect of the visual stimulus, is tightly correlated to the external timescale of the visual stimulus, i.e., the speed with which the visual image on the retina changes.

Visualizing how the brain encodes information

The institute’s scientists have developed a special visualization technique for how multiple neurons fire spikes together to encode information, by representing the identity of firing patterns of multiple neurons with color sequences.

More news