Hello ~ I am Haimeng Zhao (赵海萌), an undergraduate of Physics and Mathematics at Zhili College, Tsinghua University. I’m deeply fascinated by how the universe works and how we can possibly understand it. So I started working in the intersection of Physics and Artificial Intelligence.

Specifically, my interest comes in two folds:

- the foundation, application and interplay of quantum physics, information theory and statistics;
- (quantum) machine learning for physics and physics for machine learning.

I’m currently an undergrad research fellow in John Preskill’s group at IQIM, Caltech. Before that I was an exchange student at EPFL, Switzerland, where I worked with Giuseppe Carleo and Filippo Vicentini on neural quantum states. In my sophomore year, I worked on AI for Astronomy with Wei Zhu at Tsinghua.

**I am joining Caltech as a Physics graduate student in 2024 Fall!**

- Quantum Information, Statistics & Learning Theory
- AI for Science, especially Physics & Astrophysics
- Quantum Many-body Physics
- Generative Learning

Undergraduate, 2020 - 2024(expected)

Physics, Tsinghua University

Outstanding Graduate, 2020

Shanghai High School

High performance scientific computing with Python & C++.

Differentiable programming with JAX, PyTorch (6 years) & TensorFlow.

(Quantum) Learning Theory,

Variational Quantum Algorithms,

Generaitive Learning,

Neural Differential Equations,

(Quantum) Monte Carlo,

Tensor Network, DFT,

Neural Quantum State

(2023).
(2023).

Non-IID Quantum Federated Learning with One-shot Communication Complexity.
Quantum Machine Intelligence, 5(1), 3.

(2023).
MAGIC: Microlensing Analysis Guided by Intelligent Computation.
The Astronomical Journal, 164(5), 192.

(2022).
(2022).
Introduced control variates to control gradient variance and significantly reduce sample complexity.
Conducted extensive numerical & theoretical studies to understand different sample complexity behavior.
Benchmarked different tomography methods and propose to design quantum-resource-efficient NQSs.

Proved the key theorem in the paper, which enables one to merge partial classifiers trained in different nodes into a global one without leaking private data.

Introduced U-Net and neural controlled differential equations to parameter estimation of microlensing.
Developed a machine learning framework for irregular astronomical time series, listed on NASA EMAC.
Accelerate microlensing analysis by 5 orders of magnitude and successfully applied to real events for the first time.

Proposed and studied the non-IID quagmire in quantum federated learning, theoretically and numerically.
Extended Liu et al. to a quantum algorithm. Conducted extensive numerics to show its robustness and efficiency.

Led a team that developed a simulation & machine learning pipeline to promote neutrino energy detection precision, a key step towards understanding the neutrino mass ordering problem.

Introduced a pruning method originally used in neural architecture search to the field of lossy image compression. Achieved the state-of-the-art performance with much simpler training procedure.