This Week's Discoveries | 7 March 2017
- Paul Baireuther
- Roberto Maiolino
- 7 maart 2017
Niels Bohrweg 2
2333 CA Leiden
- De Sitterzaal
Quantum Error Correction with Deep Neural Networks
Paul Baireuther (LION) is a PhD student with Carlo Beenakker at the Institute-Lorentz of LION. His thesis defence is scheduled for 26 April.
'Artificial intelligence is quickly fulfilling its promise for vast improvements across all areas of our lives. Recent breakthroughs in deep machine learning have catapulted it from academia to the forefront of the big data industry. Beyond the feats of beating some of the world's best players in games such as Jeopardy and Go, a plethora of commercially viable applications have been developed, and are in the process of being further perfected. Examples include voice-to-text conversion, translation, automatic image labeling, and self-driving cars.
Another technology that aspires to revolutionize the way we do computation is the quantum computer. Here, the classical bit is replaced by the quantum bit (qubit). However, these qubits are very sensitive to noise and lose their information too quickly to do sophisticated calculations. A promising approach to solving this challenge is to use several physical qubits to encode an effective qubit. A part of the system can then be repeatedly measured without changing the state of the effective qubit. The so-gained information is then interpreted by a `classical decoder' to detect and correct errors. However, this decoding is a highly non-trivial task, and the most common approaches using the blossom algorithm of Edmonds are complicated and difficult to implement.
In our research we try to solve the decoding problem using a deep neural network. Using this approach we hope to create a multi-purpose decoding architecture that might overcome current limitations of even the most sophisticated decoders. In this talk I will give an introduction to neural networks and how they can be used for quantum error correction.'
Second Lecture, Lorentz Center highlight
The role of massive galactic outflows in galaxy evolution
Roberto Maiolino (University of Cambridge) is professor in Experimental Astrophysics and director of the Kavli Institute for Cosmology at the University of Cambridge (UK). He is active in multiple areas of observational extragalactic astrophysics. He is one of the participant of the workshop “Observations and Theory of Quasar Outflows” that is being held in the Lorentz Center from 6 Mar 2017 through 9 Mar 2017
'Both active star formation and supermassive black hole accretion can drive very energetic winds that can expel large amount of gas out of galaxies. I will present some recent observational results illustrating the dramatic effect that such winds can have on galaxy evolution. In particular, in some cases these energetic winds can clean galaxies of most of their gas content. This may result in a significant suppression of star formation in galaxies. However, I will also show that the gas expelled out of galaxies is not simply lost, it can also undergo compression, fragementation and finally form new stars. This is a completely new mode of star formation, occurring outside galactic discs, and resulting in stars with very high velocities, potentially even escaping the galaxy. I will shortly discuss the potential implications of this new mode of star formation.'