Research Vision

CINDLab’s research seeks to uncover universal computational principles that govern both biological and artificial systems of intelligence. Our goal is to understand how complex neural populations encode information, maintain stability, and flexibly adapt to changing environments. By integrating data-driven analysis, theoretical modeling, and AI-based simulation, we aim to identify the mechanisms through which intelligent behavior emerges from dynamic, noisy, and distributed neural systems.

Our studies focus on several foundational questions:

• How do distributed neural circuits achieve stable cognition amid intrinsic and environmental noise?

• What statistical laws describe the variability, synchronization, and coordination of neuronal activity across brain regions?

• How can deep learning architectures be constrained by biological realism to achieve robustness, interpretability, and adaptability?

By bridging quantitative theories of neural computation with machine learning frameworks, CINDLab seeks to define a unified language of intelligence—one that connects microscopic neural mechanisms to macroscopic cognitive functions. Through collaborative, cross-disciplinary research, we strive to translate biological insights into next-generation AI systems while using AI itself as a tool to decode the dynamics of the brain.

Methodological Innovation

• Computational Modeling of Neural Systems
  We construct and analyze biophysically inspired recurrent and spiking network models to reproduce experimentally observed dynamics in cortical regions such as the prefrontal cortex (PFC) and posterior parietal cortex (PPC). These models reveal mechanisms underlying working memory, attentional gating, and decision variability.

• Statistical and Mathematical Modeling
  The lab develops advanced probability distributions and stochastic process models for analyzing complex, non-Gaussian, or heavy-tailed data from biological and behavioral systems. This includes new distribution families for reliability data and neural spike statistics, as well as latent-variable frameworks for identifying hidden structure in large datasets.

• AI and Deep Learning Integration
  We apply and extend deep neural architectures—including recurrent neural networks (RNNs), convolutional neural networks (CNNs), and attention mechanisms—to model brain-like computations. Our goal is to identify convergent principles between machine learning optimization and biological adaptation, thereby improving both the interpretability of AI and the realism of computational neuroscience models.