Our research focuses on the theoretical model and computational simulation of functional structures of the brain, to find underlying mechanisms of how our brain performs various functions. Currently, we are interested in various functional maps in visual cortex, trying to find a unified theory of developmental mechanism of those maps. We also study how neurons interact with each other within these networks, and how they encode/decode various features of sensory information. For this, we examine nonlinear neural activities in a large neural network scale, or in a system level.
Our research is mostly performed based on theoretical modeling and computational simulation methods, using Matlab, GENESIS, and other programming languages, but we also collaborate with other experimental groups in the related fields. Our research helps us better understand how the functional structures in brain develops and operates for the optimal process of information, and in the long run, this may enable us to try designing some brain-inspired applications such as a neuro-computer or a task-specific artificial intelligence. The main directions of our current work are described below
1. Theoretical Model: Functional Maps in Sensory Cortex
In the primary visual cortex of higher mammals, neurons respond selectively to the various features of visual stimuli and are spatially organized by their selectivity, forming various types of functional maps. Explaining the origin and role of these maps is crucial for understanding sensory systems and has been considered one of the most interesting issues in neuroscience. Currently, however, very little is known about the mechanisms by which these maps arise, how they are modified during development, and what function they may play.
Our research interest is in the neural circuits of the functional architecture of the brain. Specifically, our research focuses on the dynamics of the functional organization of the neural networks in visual system. 
We are interested in how neurons interact with each other within a large neural circuit and how this affects the structure of functional maps and their response to visual stimuli, which will provide better understanding of the working mechanism of brain.
Neurons in the brain have plasticity in their structure and connectivity. This feature is particularly important during critical periods of development, in which the functional maps of the brain are refined and modified. Our work is to explain the features of various functional maps in young and adult brains. This will help us to find the probable advantage of biological circuits over the silicon-based. Computational simulation is an effective tool to test theoretical models for this type of task and useful for making predictions and planning the direction of experimental work.
2. Computational Simulation of Large Neural Networks: Learning and Memory
How does the human brain perform various functions? This is probably one of the most intriguing and challenging questions of the current time. In fact, understanding the working mechanism of the brain is a difficult task because of the large number of constituent neurons and extremely complicated neural interactions in the brain circuits. Moreover, in most cases, important functional features of neural system can only be observed in a system level, thus must be studied in a large-scale neural population. Therefore, in experimental brain studies, exploring the neural functions in detail is a perplexing problem because of the difficulty in controlling and measuring many relevant variables simultaneously.
In this case, theoretical modeling and computational simulation can provide insights into the task. In the past decades, computing power of the machine has been increased enormously, which enabled us to perform realistic computer simulations of the complicated neural networks.
We design a computer simulation using a model feedforward network where various types of learning and memory models are implemented and tested. We examine the performance of the system to study underlying process of human memory. For instance, we are interested in how “sustainability” and “appendability” of memory can be regulated or controlled by neural parameters, to understand differences between a short-term and a long-term memory.
Our research aims to find a simple but powerful theory that the various features of memory can be readily explained. Our simulation approach can shed light on the research of how memory is formed, erased, and can be controlled.
3. Human psychophysics experiment : Understanding visual perception
When we see an ambiguous visual stimulus such as Necker cube, our perceived state switches periodically between two possible interpretations. This phenomenon, called bistable perception, is critical to our understanding of the mechanisms underlying sensory perception and information processing in the brain because it is a consequence of dynamic interpretation of incomplete information, without any change in external stimuli. However, the question of how one’s perceptual state flips to another, or even why bistable perception occurs, has not yet been fully answered. Under various environment, we conduct human psychophysics experiment with bistable stimuli and analyze the human response using computational model to reveal the functional implication of sensory perception.
4. Neural Activity Data Analysis
In neural imaging data, various types of spatio-temporal neural network activity patterns are observed that may reflect important dynamic features of information processing in the brain. However, conventional method analyzed imaging data with much information loss and unwanted dimensionality reduction due to the complexity and nonlinearity of activity patterns. Thus, we developed a computational analysis method for the classification of activity patterns in two aspects.
