Yicong (Alan) ZHENG

Research Overview

My research sits at the intersection of computational neuroscience and artificial intelligence, where I develop biologically-inspired models to understand how the brain processes information and how these insights can advance AI systems. At Astera Institute, I focus on bridging the gap between neuroscience and machine learning by creating computational frameworks that are both scientifically rigorous and technologically innovative.

Research Areas

AI and Machine Learning

LLM agents and RL
I apply reinforcement learning to train multi-turn LLM agents to perform tasks in a constrained environment. I am also adding memory modules to these agents to enable them to continously learn from their experiences.

Biologically-Plausible Neural Networks
I develop neural network architectures inspired by brain circuits, incorporating biological constraints like bidirectional connectivities between neurons, realistic connectivity patterns, and error-driven plasticity. These models achieve competitive performance while maintaining interpretability and biological realism.

Continual Learning and Memory Systems
Drawing from neuroscience research on hippocampal-neocortical interactions, I design AI systems capable of continual learning to minimize the impact of catastrophic forgetting. This work bridges computational theories of memory consolidation with practical machine learning applications.

Computational Cognitive Neuroscience

Memory Systems and Error-Driven Learning
I investigate how the brain forms, stores, and retrieves memories through computational modeling and neuroimaging. My work challenges traditional Hebbian learning theories by proposing that the hippocampus uses error-driven learning mechanisms similar to the neocortex. This research has fundamental implications for understanding memory disorders and developing more brain-like AI systems.

Neural Oscillations and Temporal Dynamics
Using EEG/fMRI and computational models, I study how neural oscillations (theta, alpha, beta rhythms) coordinate memory processes across brain networks. This work reveals how disrupted oscillatory patterns contribute to cognitive deficits in psychiatric disorders, particularly schizophrenia.

Spatial Cognition and Navigation
Through virtual reality experiments combined with neuroimaging, I examine how the brain represents spatial information, investigating the neural mechanisms underlying spatial memory and exploring how different brain regions encode structural versus content information during navigation.