III. System Level · Systems
The system level studies how the nervous system implements sensation, movement, memory, and reward-based learning. From orientation selectivity in visual cortex to cerebellar motor learning, from associative memory in Hopfield networks to reward prediction error in the dopamine system — each system has its own computational strategy.
Visual System
— Gabor Filters and Orientation TuningV1 simple cells have receptive fields described by Gabor functions. Tuning curves characterize each cell's preference for different orientations.
Auditory System
— Frequency Tuning and Cochlear TonotopyA γ-tone filter bank mimics the basilar membrane's frequency analysis. Characteristic frequencies are arranged logarithmically along the membrane (tonotopy).
Motor System
— VOR Gain Adaptation and Cerebellar LearningThe vestibulo-ocular reflex gain is learned through error-driven cerebellar learning (Marr–Albus rule).
Associative Memory
— Hopfield Networkw_ij = 1/N·Σξᵢξⱼ. Asynchronous updates monotonically decrease the energy, restoring corrupted patterns to stored memories.
Reward Learning
— Rescorla–Wagner and TD(λ)ΔV = α·(λ-V). Reward prediction error drives learning. The TD model reproduces the blocking effect and dopamine-like signals.