Professor Ard Louis gives a basic introduction to deep learning for physicists and addresses a few questions such as: Is the hype around deep learning justified, or are we about to hit some fundamental limitations? In less than ten years, machine learning techniques based on deep neural networks have moved from relative obscurity to central stage in the AI industry. Large firms such as Google and Facebook are pouring billions into research and development of these new technologies. The use of deep learning in physics is also growing exponentially. Can physics help us understand why deep learning works so well? And conversely: How can deep learning provide new insight into the world around us?

An Introduction to deep learning

Learn about quantum mechanics, black holes, dark matter, plasma, particle accelerators, the Large Hadron Collider and other key Theoretical Physics topics. The Rudolf Peierls Centre for Theoretical Physics holds morning sessions consisting of three talks, pitched to explain an area of our research to an audience familiar with physics at about second-year undergraduate level.

Education

Training

Language Courses

Higher Education

K-12

Educational Technology

In this talk, Benedikt Placke introduces QEC and explains how the unique interplay between the classical and the quantum world enables us to efficiently correct errors effecting such systems.  Quantum computing is a new model of computation that holds the promise of significantly improved performance over classical computing for some problems of interest. However, by its very nature quantum computers are sensitive to disturbance by external noise, most likely necessitating the use quantum error correction (QEC) for useful application.

Furthermore, Benedikt Placke comments on the deep connection between QEC and questions in condensed matter physics.

The Miracle of Quantum Error Correction

In this talk Alessio Lerose discusses the seminal idea of simulating Nature via a controllable quantum system rather than a classical computer. He discusses recent advances that brought us closer to the ultimate goal of a universal quantum simulator. Since their birth computers proved invaluable tools for physics research. Quantum mechanics, however, fundamentally challenges the possibility for computers to simulate dynamics of matter. In fact, solving the quantum-mechanical law of motion requires to account for contributions of all possible joint configuration histories of all constituents of a system: a task that quickly becomes unbearable for any imaginable computer. Our understanding of complex phenomena involving important quantum-mechanical effects, such as chemical reactions, high-temperature superconducting materials, as well as the primordial universe evolution, is obstructed by this fundamental technological limitation.

Simulating physics beyond computer power

Jasmine Brewer covers recent progress on studying the properties of the quark-gluon plasma, and describe how we can capitalize on lessons learned from high-energy physics to provide new insights on this novel material. Quarks and gluons are the fundamental constituents of all matter in the universe, but they have the unique property that they are always confined inside hadrons. The only situation in which quarks and gluons are deconfined is in extremely high-energy collisions of heavy nuclei, where the temperature is so high that nuclei “melt” into a new phase of matter called the quark-gluon plasma. This exotic state of matter provides a gateway to study the rich many-body physics of free quarks and gluons, including their rapid thermalization to form the most perfect liquid ever observed.

A liquid of quarks and gluons

Dr Yonadav Barry Ginat - Possible sources for the gravitational wave background The detection of gravitational waves from the coalescence of black holes has opened a new window for astronomy. Besides individual mergers, one can study the stochastic gravitational-wave background, i.e. the sum of all gravitational waves arriving at Earth, which are not from resolved sources. In this talk I will give an overview of the current predictions for this background, over a range of frequencies -- from binary neutron stars at 100 Hz to the mergers of super-massive black holes at 10^(-8) Hz, and even further to primordial gravitational waves generated during inflation. Of these, none have so far been detected, save for a signal consistent with a background from super-massive black hole coalescences. I will touch on how background sources are modelled, and on how these can be used to extend our understanding of physics.

Possible sources for the gravitational wave background

Prof Bence Kocsis - Searching for the origin of black hole mergers in the Universe with gravitational waves The direct detection of gravitational waves by LIGO and VIRGO and pulsar timing arrays has recently opened a new window to observe the Universe. We can now detect objects which are completely invisible in traditional electromagnetic surveys including black holes and possibly dark matter. The observations show a very frequent rate of black hole mergers in the Universe with unexpected properties. In this talk I will review the astrophysical processes that may be responsible for the formation of the observed events. I will show that the standard astrophysical merger pathways are already in tension with LIGO/VIRGO observations. New ideas may be needed to explain the origin of the detected sources. I will discuss several exotic possibilities including the hypothesis that if dark matter is in part made up of black holes in galaxies they may contribute to the observed events or the possibility that stellar mass black holes may be teeming around supermassive black holes at the centres of galaxies, which may be a possible sight to produce gravitational wave events.

An Introduction to deep learning

Download our free app to listen on your phone