Searching for the symmetry, mesoscopic structure and order parameter in the brain.

I'm looking for Fall2027 Neurotheory PhD position in the US.

Research Interests: Statistical Mechanics for Networks of Real Neurons. Theoretical Neuroscience.

Now I'm working with Prof. Florian Engert at Harvard, developing physics inspired data analysis methods, to simplify and integrate the structural (EM connectome) and functional (calcium imaing) data. (02/2026)

Now I am focusing on using maximum entropy model, renormalization group based coarse-graining and other statistical mechanics inspired methods to analyze neural data, including mouse V1 and zebrafish calcium imaing and Flywire connectome data. (12/2025)
PS: I'm also interested in analytical methods from statistical mechanics and mathematics, such as different mean-field methods. But in my current stage, I'm still searching for (or developing) interesting neural dynamical models to study, possibly containing multi-scale interactions like spiking, electric field and ephaptic coupling. This is also why I work on neural data analysis now: I believe that real neuronal activity contains unknown (and essential!) structures that are not yet captured by existing neural dynamical models.

Generally speaking, my current and near-future research focuses on physics inspired phenomenological modeling of real neural circuits, positioned somewhere between neural dynamical modeling and mainstream statistical machine learning. I am not interested in (faithfully) reproducing time-varying neural activity patterns — which is the traditional neural dynamics approach. At the same time, my work also differs from standard machine learning approaches based mainly on direct data fitting. Instead, my modeling is guided more by physical insights.

If you are interested, you can know my recent work through this figure. I also wrote parts of my research proposal and long-term research goals for graduate school applications. However, because some institutions imposed strict word limits, I had to keep them relatively concise. (12/2025)

------------------------------------------------------------------------------------------------------

Due to PhD applications, the notes on this page (including topics such as path integral and Koopman theory) have not yet been fully uploaded. (August 1, 2024)

A metaphor for my planned PhD research in neuroscience:

As humans, we are eager to become gods. We live on the ground, while gods live in the sky. The mainstream approach is to search for clouds in the sky that close to the god. We take those clouds as targets and strive to elevate ourselves toward them (like standing on increasingly higher heels), hoping one day to reach them. We repeat this process, constantly pursuing ever higher clouds, believing that could ultimately approach gods reside at the very top. However, this is not the path I wish to follow.

In my view, for humans to become gods, the goal is not to strengthen our own power or raise ourselves ever higher, but rather to bring the gods down to our level—to strip away the decorative exterior that surrounds them until they become something within human reach (something that could be touched with only a jump). To realize this vision, my plan is as follows:

Describe the gods — this naturally includes studying the clouds.
Pull the gods down from the sky, as low as possible.
Describe the “god-like beings” that already exist among us, in order to understand what characteristics they possess.
Using the insights gained from the previous steps, take a major step toward becoming gods ourselves.

Throughout this process, we may occasionally need to elavate ourselves, so that our observations and actions become wiser and more effective.

Note: Since I am currently in the period of the application process, I cannot yet fully explain the precise meaning of this metaphor. Once the application process is over, I would be happy to elaborate on it in detail here.

Phd period proposal (Outdated):
Stage 1: Construct coarse-grained theoretical frameworks and uncover features of neural design principle with the help of phenomenological models, i.e. Maximum Entropy Model.
Stage 2: Construct Lower-dimensional theories of neural design principle with the help of heuristic models and finding corresponding brain mechanisms to support such coarse-grained computation. (candidate: electric field and ephaptic coupling)

I dream of achieving the following goals by studying and researching neurotheory and believe that such coarse grained theories are as important as some biological experiments.

What is the definition (or correct level) of "understanding" the brain? Multi-scale theories? Simulating all the detail in the brain? Unfortunately, many neuroscience study unawarely (so called causal chain study and classical 3-steps: analyzing, modeling and experiment) imply this idea. I proposed a new possibility: transferring the consciousness. Although it is still not operationalizable at the moment, it at least encourages a top-down research approach — unless one simply believes that AI models will suddenly emerge consciousness someday. Based on this idea, I.....

Here are some older views that have not yet been updated:
I believe there are lower dimensional structures in the brain(at least in some brain regions) relative to the number of neurons (although it would still be high dimensional). Compared to traditional linear-dimensional reduction method(latent factors), there should be some new spaces(not state space!) contain such lower dimensional structure. Here are some related work(based on different context or views): MEM, learnability, Control theory(Perturbation). Considering that these conjectures/conclusions will be closely related to what I consider to be two milestones in neuroscience - the simultaneous recording of all neuronal activity in the whole brain at a certain resolution, and the recording of all the biological components in the CNS - the best place to test the theory, at least at this stage, will be C. elegans and Larval Zebrafish and other small animals.

Mid-term goal: control theory of the nervous system, which currently seems to be best studied in small animals (C. elegans and Larval Zebrafish). This includes, but is not limited to, studying the sensitivity and causal-functional connectivity of neural networks and their relationship to network function. And the discovery of generalized low-dimensional structures in the neural circuits.

Ultimate goal: simulate (transferred from the biological brain) personal consciousness in silico.

While lately I am very interested in some specific topics in theoretical neuroscience: Cerebellum-like Structures, Geometry of Perceptual Manifolds, Clustering Principles of Neural Activity, Bridging Neural Manifolds and Neural Circuits, and how the new neurophysiological phenomena observed at molecular and single cellular level in experiment effects macroscopic neuronal networks dynamics.

Here's a detailed version of my research interest (informal) on July 21st

Experience:

Research Scholar Harvard University Supervisor: Florian Engert 2026.2-present
Research Intern University of Washington Supervisor: Adrienne Fairhall and Leenoy Meshulam 2023.7-2025.11
Research Student Shanghai JiaoTong Univerisity Supervisor: Songting Li and Douglas Zhou 2022.2-2023.1
Research Assistant Shanghai JiaoTong Univerisity Supervisor: Ru-Yuan Zhang 2021.8-2022.1
Research Assistant Xi'an Jiaotong-Liverpool University Supervisor: Kaizhu Huang 2020.6-2021.6

Zhihu(Quora Chinese version)

Twitter
https://twitter.com/ChongwenWang