[Seminar] Neural Emergence and Plasticity of Color Dimensionality in Human Visual Perception

Dear All,

The Neural Computation Unit (Doya Unit) would like to invite you to the following seminar:

Speaker: Atsunobu Kotani (University of California, Berkeley)

Title: Neural Emergence and Plasticity of Color Dimensionality in Human Visual Perception

Abstract: Imagine you are in the Louvre, looking at the Mona Lisa. You see a woman with a faint smile, yet your brain never receives a colored picture: the optic nerve transmits only a rapidly changing pattern of binary spikes. This dissertation asks how the visual cortex constructs color perception from these signals alone, and whether color dimensionality—the number of primaries needed to match any color—can expand when a new cone signal becomes available in adulthood.

In this talk, I will present two central parts of this work. First, I will introduce a computational theory based on self-supervised predictive learning, solely from optic nerve signals. Without reference images, cone labels, or a predefined color space, the model learns color dimensionality that matches the number of cone types; when a third cone type is added after dichromatic development, continued learning reorganizes the model from two- to three-dimensional color vision. Second, I will present a human pilot using Oz+, which supplies an adult dichromat with a virtual third-cone signal through Oz Vision, a recently developed adaptive-optics system that targets thousands of mapped cones. Under Oz+, measured color dimensionality increased from two to three, and color discrimination and visual search improved. Although Oz+ reproduced the mottled cone-activation pattern underlying a trichromat’s view of a uniform red or green surface, the participant perceived it as much closer to uniform—yet did not report seeing a previously unseen color. Together, these results suggest that functional use of a new sensory signal and the subjective experience of new colors can dissociate.

Zoom URL:
https://oist.zoom.us/j/99095830326?pwd=hTzEEe4ns2eIeGbW1RaiDt87Tzb4do.1
Meeting ID: 990 9583 0326
Passcode: 180843

We hope to see many of you at the seminar.

Sincerely,
Neural Computation Unit
Contact: ncus@oist.jp