Research

Emergence of Cognition Functional Circuits Neuro-AI Learning and Memory Collaborative Topics

We conduct research using tools from computational and theoretical neuroscience, to comprehend a wide range of topics, focusing on: 1) theoretical modeling and computational simulation, 2) computational analysis of data from animal experiments, 3) development of neuro-AI models, and 4) human psychophysics.

Emergence of Cognitive Intelligence in Natural and Artificial Neural Networks

Can brain functions be developed solely through training, like those in artificial neural networks? How do “innate functions” emerge in the brain prior to training?

Our research using an artificial neural network model shows that cognitive functions like quantity estimation and object perception can emerge spontaneously in randomly initialized deep neural networks without learning, offering insights into the mechanisms of innate cognitive functions.

This raises new questions: what other types of functions can emerge spontaneously in neural networks? How many cognitive functions are innate, and how can we predict and find them? Can we create a complete table of all innate functions present in the brain?

Spontaneous Emergence of Functional Circuits in the Brain

“Tabula rasa” vs. “Ex nihilo” - A Biological Strategy Reveals How Efficient Brain Circuitry Develops Spontaneously

Our researches have explained how the regularly structured functional circuits in the brain, such as topographic cortical maps, could arise spontaneously to efficiently process sensory information.

This raises additional questions: How do these spontaneously formed structures develop over time, and how do they contribute to the emergence of more complex cognitive functions? Could the principles of this emergence be applied to the functionality of artificial neural networks?

Biological Factors Predict Distinct Cortical Organizations across Mammalian Species

We have shown how visual circuits develop differently across mammalian species, identifying key biological factors like the retino-cortical mapping ratio that predict distinct cortical organizations. These findings resolve a long-standing puzzle about sensory system architecture and demonstrate that evolutionary variation in biological parameters can lead to distinct functional circuits without species-specific developmental mechanisms.

This raises the question: Can we infer the relationship between brain structure differences and functional variations across species? Through this understanding, can we predict the direction of brain evolution and propose the brain structure of new species?

Development of algorithms for Neuro-AI

Comparative Study between Biological and Artificial Neural Networks

Our research shows that the biological brain organizes visual circuits more efficiently than artificial neural networks. The brain uses fewer, shallower hierarchical networks for visual object recognition, unlike deep neural networks in AI. We aim to explore anatomical structures in the brain that enable this efficiency but are absent in artificial networks.

By comparing brain functions with AI models, can we identify key differences that could improve AI systems and offer new directions for biology-inspired computer vision?

Robust Artificial Visual Intelligence Model of Biological Visual System

Our research examines the limitations of traditional deep neural networks, which struggle with domain-specific biases and visual variations, unlike the adaptable biological brain. Our models maintain robust performance across environmental shifts, aligning with human visual perception. This biologically inspired approach aims to enhance AI models for reliable performance in diverse conditions.

The next question is: How can we further mimic more complex aspects of biological vision? What additional brain mechanisms can be incorporated to enhance AI's robustness across even more diverse and dynamic environments?

Random Pretraining of AI Inspired by Developmental Neuroscience

Unlike AI systems that rely solely on data for training, the biological brain begins learning through random neural activity before sensory input. Inspired by developmental neuroscience, we propose random noise pretraining for AI models. Our research shows that this approach improves learning efficiency and generalization in biologically plausible networks using local learning, and enhances uncertainty calibration in backpropagation-based networks. This highlights how developmental neuroscience insights can address resource efficiency and reliability challenges in AI.

How can these neuroscience-inspired insights contribute further to solving other challenges in AI, such as adaptability, energy efficiency, and robustness?

Learning and Memory: Flexible and Stable learning in the Brain and AI

Unlike computers, which suffer from catastrophic forgetting, the human brain supports lifelong learning, integrating new knowledge while preserving past experiences. How does this balance emerge, and can AI achieve the same adaptability?

Our research examines memory functions in both the human brain and AI. We found that human working memory follows the serial-position effect, where people remember the first and last items better than those in the middle. This may result from the brain using both stable and flexible strategies to preserve previous memories while supporting new learning. At the neural circuit level, synaptic adjustments over time determine whether memories remain stable or flexible, highlighting the brain’s balance between retention and adaptability.

Can we apply these findings to artificial neural networks to enable them to exhibit memory characteristics similar to humans?

Social Behavior in Multi-Agent Systems: From humans and primates to birds, whales, and insects, organisms form societies where behaviors like culture, collective action, and group intelligence emerge. What fundamental mechanisms drive these behaviors?

Our research explores how simple learning rules at the individual level can lead to social behaviors in multi-agent systems. Inspired by animal group behaviors, we investigate whether randomness in learning, like spontaneous deviations or unbiased imitation, contributes to social behavior formation. Through multi-agent simulations, we examine if random copying can drive group adaptation and behavioral optimization. We also explore how stochasticity in individual choices enhances coordination, showing that occasional deviations improve system efficiency. This work aims to uncover how simple mechanisms can lead to social behaviors in both biological and AI-agent systems.

Then, can we find principles for optimizing multi-agent environments by studying the collective behaviors exhibited by various animal species in nature?

Collaborative Topics

We collaborate with leading experimental research groups. Our work has contributed key insights into brain function and neurological disorders, including studies on various brain disease models such as Parkinson's disease, autism spectrum disorder, and epilepsy from somatic mutations. Our expertise in computational analysis has not only validated experimental findings but also provided theoretical frameworks to better understand these observations. Through these collaborations, we integrate computational approaches with experimental data to advance our understanding of neural systems.