-

The next generation of artificial retinas will rely on electronics capable of reproducing the natural pattern of activation of the different neuron types with single-cell and single-spike resolution. They will need to calibrate the stimulation parameters based on the recorded neural responses in a closed-loop manner. However, due to the complex electrode-tissue interface, a stimulation artifact corrupts the recorded neural activity at different electrodes, severely limiting the calibration process. Here, we propose a new mechanism suitable for implantable massively parallel interfaces to mitigate artifacts using linear models of the electrode-tissue interface and convex optimization algorithms.

An AI scientist Richard Feynman famously said "What I cannot create, I do not understand", referring to the principle of gaining a deep understanding of a matter by deriving it from a blank piece of paper and just the fundamental concepts. Scientists and engineers follow this principle when they distil natural phenomena into concise mathematic equations and build test prototypes of complex machines. But doing so requires extensive work, begging the question if we can automate it? Scientists at TU Delft are exploring how artificial intelligence can help us discover explanations of natural phenomena around us.

Quantum Conversaties: Een Sprong Voorwaarts met Nieuwe Qubits In kwantumcomputing gebruiken we qubits, net zoals huidige computers bits gebruiken. De fluxonium qubit is een nieuwere soort die kwantumcomputers zou kunnen verbeteren. Maar er is een uitdaging: ervoor zorgen dat deze qubits efficiënt "aantekeningen vergelijken" of de juiste data delen, wat wij pariteitsmetingen noemen. Ons team is pionier in een methode om dit proces beter te maken voor deze nieuwe qubits. Succes hierin zou de kwantumcomputing kunnen revolutioneren, wat leidt tot nieuwe wetenschappelijke ontdekkingen.

Network algorithms are ubiquitous: you use them for example when finding the shortest route home. In this project, insights from combinatorics are used to find better algorithms, bringing fundamental mathematics a step in the direction of real-life applications. The project includes reconstructing properties from as few questions as possible (``query reconstruction') and recognizing redundant steps in sorting algorithms (``slow sorting'). The latter has implications for routing algorithms.