Search Talks

Simulations that numerically solve Einstein's equations are the only means to accurately predict the outcome of the merger of two black holes. The most important outputs from these simulations are the gravitational waveforms, and the mass and spin of the final black hole formed after the merger. The waveforms are used in extracting astrophysical information from detections, while the final mass and spin are used in testing general relativity. Unfortunately, these simulations are too expensive for direct use in data analysis; each simulation can take a month on a supercomputer. Surrogate modeling is a data-driven approach to modeling, that uses machine learning like techniques to interpolate between hundreds of existing simulations. In this talk, I will discuss some recent developments in surrogate modeling, including a new 7-dimensional model that fully captures the effects of precession; the wobbling of the orbit caused when the black holes’ spin axes aren’t perpendicular to the plane of the orbit. This model reproduces the waveform as well as the final black hole's mass and spin as accurately as the simulations themselves, while taking only a fraction of a second to evaluate on a laptop.

Machine learning has led to recent advancements in image processing, language translation, finance, robotics, musical and visual arts, and medical diagnosis. In this session, we will explore how machine learning can be applied within fields of physics. We will introduce fundamental concepts in machine learning such a neural networks and supervised vs. unsupervised learning, and then proceed to learn to use tools from Python's TensorFlow library.

Bring your laptop. You can attend remotely via Zoom <https://zoom.us/j/154009181>.

Quantum complexity is a notion characterizing the universality of the entanglement arising from a quantum evolution. A universal evolution will result in a complex entanglement. At the same time, this also corresponds to small fluctuations and to unlearnability from the point of view of machine learning. All these aspects are connected to the different features of k-designs, which are under-samplings of the Hilbert space. 

We study the transition in complexity due to the doping of a quantum circuit by universal gates and show that the transition to complex entanglement can be obtained by just a single gate. 
These results are relevant for the notions of scrambling, quantum chaos, OTOCs and operator spreading.

We conjecture that the transition to 4−design, W-D and unlearnability are one and the same.

 Python is a widely used programming language for scientific
computing, data analysis, and machine learning. This tutorial will
feature an introduction for those with little or no background.

Large galaxy surveys have dramatically improved our understanding of astrophysics and cosmology in the high-redshift universe, but they are fundamentally limited by the need to integrate long enough to detect each individual source. Line intensity mapping has recently arisen as a powerful alternative to these surveys, offering access to fainter sources and larger volumes than conventional techniques. There has been a surge of experimental interest in this technique, with surveys planned or in progress across the electromagnetic spectrum. In this talk, I will describe the wide variety of science which we will obtain from these experiments in the next few years and illustrate the methods by which we can go from maps of confused line emission to useful astrophysics. I will show how intensity maps can give new insights into topics ranging from star formation to the high-redshift ISM to the Hubble constant tension. I will further discuss the utility of combining intensity maps with conventional surveys, both for systematics control and for studying processes like AGN feedback. I will close with a discussion of how modern machine learning methods can be used to further extend what we can learn from these surveys.

Self-learning Monte Carlo (SLMC) method is a general-purpose numerical method to simulate many-body systems. SLMC can efficiently cure the critical slowing down in both bosonic and fermionic systems. Moreover, for fermionic systems, SLMC can generally reduce the computational complexity and speed up simulations even away from the critical points. For example, SLMC is more than 1000 times faster than the conventional method for the double exchange model in 8*8*8 cubic lattice. In addition, SLMC also provides a general framework to naturally integrate the advanced machine learning techniques into Monte Carlo. In this talk, I will give an introduction about the background, basic idea and the design principle of SLMC. Later, I will explicitly show how to use SLMC and its great accelerations in classic systems, free fermions coupled with classical spins systems, and interacting fermion systems. At the end, I will talk our recent developed a new type of neural network, which is motivated by the mean field theory and could be used to represent many physical Hamiltonians. We believe the new structural network would be highly efficient to identify different phases and accelerate the numerical simulations for even complex models.