Heng Zhang

Heng Zhang

Postdoctoral Researcher at IRCN, University of Tokyo

Developing bio-inspired AI systems for computer vision and NLP through predictive processing and self-organization principles.

Background

I am a postdoctoral researcher at the International Research Center for Neurointelligence (IRCN), University of Tokyo, in collaboration with KTH Royal Institute of Technology. My research develops bio-inspired AI systems for computer vision and natural language processing by drawing on principles of predictive processing and self-organization observed in biological systems.

I focus on creating adaptive and robust learning frameworks that excel in dynamic, uncertain environments. My work spans unsupervised learning, dynamical systems, neural representations, and sequence learning. Through computational modeling of brain functions—particularly prediction and complex sequence processing—I aim to develop artificial intelligence that reflects the efficiency and resilience of biological cognition.

Specialties

Deep Learning Computer Vision NLP Reservoir Computing Self-Organization Dynamical Systems Unsupervised Learning Robustness PyTorch TensorFlow Python MATLAB

Research Interests

Self-Organizing Dynamical Equations

Self-Organizing Dynamical Equations (SODE)

Bio-inspired systems for complex sequential learning using self-organization and nonlinear dynamics. Enables adaptive learning without backpropagation. Read more

Robust Bio-inspired Vision Systems

Robust Bio-inspired Vision Systems

Unsupervised image segmentation systems that maintain performance under severe noise and corruption, mimicking human visual perception. Read more

Xenovert Adaptation

Xenovert: Adaptive Distribution Shift

Biologically-inspired algorithm for real-time adaptation to distribution shifts using tree-based mapping structures. Read more

Reservoir Computing

Advancing Reservoir Computing

Bio-inspired sparse recurrent neural networks leveraging high-dimensional dynamics for temporal and spatial processing. Read more

Temporal Learning Systems

Neuro-like Temporal Learning Systems

BCPNN-based attractor networks for rare-event learning in temporal data streams with modular hypercolumn architecture. Read more

Selected Works

Under Review

SyncMapV2: Robust and Adaptive Unsupervised Segmentation

Heng Zhang, Zikang Wan, Danilo Vasconcellos Vargas

IEEE Trans. Pattern Analysis and Machine Intelligence, 2026

Under Review

RECAP: Local Hebbian Prototype Learning as a Self-Organizing Readout for Reservoir Dynamics

Heng Zhang

Neurocomputing, 2026

Under Review

Self-Organizing to Learn Dynamical Hierarchies

Danilo Vasconcellos Vargas, Tham Yik Foong, Heng Zhang

Scientific Reports, 2025

Under Review

Adapting to Covariate Shift in Real-time by Encoding Trees with Motion Equations

Tham Yik Foong, Heng Zhang, Mao Po Yuan, Danilo Vasconcellos Vargas

In Submission, 2025

Journal Article

Symmetrical SyncMap for imbalanced general chunking problems

Heng Zhang, Danilo Vasconcellos Vargas

Physica D: Nonlinear Phenomena (Top 6%/9% in Mathematics/Nonlinear Dynamics), IF: 3.7, 2023

Journal Article

Magnum: Tackling high-dimensional structures with self-organization

Mao Po Yuan, Yikfoong Tham, Heng Zhang, Danilo Vasconcellos Vargas

Neurocomputing (Top 4%/7% in Cognitive Neuroscience/Computer Science), IF: 5.7, 2023

Review Article

A survey on reservoir computing and its interdisciplinary applications beyond traditional machine learning

Heng Zhang, Danilo Vasconcellos Vargas

IEEE Access (Top 8% in General Engineering), IF: 3.5, 2023 — 60+ citations

Conference

Understanding SyncMap's Dynamics and Its Self-organization Properties: A Space-time Analysis

Heng Zhang, Danilo Vasconcellos Vargas

5th Artificial Intelligence and Cloud Computing Conference (AICCC), Osaka, 2022

Conference

Comparison of Disturbance Rejection Performance between Three Types of UAV Linear Controllers

Chao Huang, Heng Zhang

2020 7th International Conference on Information Science and Control Engineering (ICISCE), IEEE, 2020

Professional Experience

April 2025 — Present

Postdoctoral Researcher

University of Tokyo — IRCN

Collaborative initiative with KTH Royal Institute of Technology. Computational modeling of brain functions including prediction and complex sequence processing for bio-inspired AI.

Oct 2024 — May 2025

Postdoctoral Researcher

Kyushu University

Developing robust and adaptive machine learning models. Building bio-inspired AI systems resistant to adversarial attacks and noise corruptions.

May 2024 — Present

System Engineer

MiraiX.org

3D open-world game development, machine learning engineering, and UI/UX design.

July 2022 — Sept 2024

Research Assistant

Kyushu University

Research assistant for projects in mathematics and information engineering. Lab assistant for stem cell culturing.

Academic Background

Oct 2021 — Sept 2024

Ph.D. in Artificial Intelligence & Information Science

Kyushu University

Thesis: "Adaptive and Robust Learning with Self-Organizing and Nonlinear Dynamics." Full-funded by SPRING program from JST.

Sept 2019 — Dec 2020

MSc. Communications & Signal Processing

University of Manchester

First-Class Honours. Outstanding Distinction grade in master's thesis on human gait classification using machine learning.

Sept 2015 — July 2019

BEng. Electronic & Information Engineering

Shenzhen University

Top 10 GPA

External Funding Contributions

UTokyo Global Activity Support Program for Young Researchers • 2026–2027

Read more

WPI-IRCN Retreat Brainstorming Awards, UTokyo • 2025–2026

Read more

The Telecommunications Advancement Foundation • 2025–2026

Read more

JSPS, MEXT Grant-in-Aid for Transformative Research Areas (A) • 2022–2026

Read more

Leading Company • 2023–2025

Read more

JSPS, Grant in Aid for Challenging Research (Exploratory) • 2022–2023

Read more

Japan Science and Technology Agency (JST) • 2021–2024

Read more

Get in Touch

Location International Research Center for Neurointelligence (IRCN)
The University of Tokyo
7-3-1 Hongo, Bunkyo-ku, Tokyo, Japan 113-0033
Email rogerzhangheng [at] gmail [dot] com