Project:
Project: Computational Modeling of Joint Encoding of Visual and Valence Cues in Amygdala
Laboratories:
James DiCarlo, Ph.D. and Rebecca Saxe, Ph.D.
Biographical Information:
Gwangsu received his BSc (2017) and PhD (2023) in Physics from Korea Advanced Institute of Science and Technology. He joined the MIT McGovern institute in 2024 as a postdoctoral researcher. During his PhD, he studied brain-like functional representations of sensory stimuli in deep neural networks and their underlying computational principles.
Current Work:
The amygdala mediates our behavioral responses to sensory inputs with positive or negative valence, and its abnormal activity can lead to psychiatric conditions like autism spectrum disorder (ASD). Our project aims to investigate how the activity of valence-coding amygdala neurons is modulated by natural visual inputs, particularly without explicit conditioning. We will develop a neural model that maps visual inputs to amygdala neuron activity, identifying specific visual features associated with valence in the primate amygdala. This research could provide key insights into how the amygdala jointly encode visual and valence information, offering potential strategies for non-invasive neural intervention to modulate abnormal amygdala activity in ASD.
Publications:
1. Kim, G., Kim, D.-K., and Jeong, H. (2024). Spontaneous emergence of rudimentary music detectors in deep neural networks, Nature Communications, 15, 148.
2. Kim, G., Jang, J., Baek, S., Song, M., & Paik, S. B. (2021). Visual number sense in untrained deep neural networks. Science Advances, 7(1), eabd6127.
Keywords:
Amygdala, Emotion, Computational neuroscience, Artificial Neural Networks