Alex Maier
We tested a timely scientific question: Does the brain (neocortex) fundamentally compute predictions (so-called predictive coding)?
Core trick
Responses to different prediction-violation types in public Allen Institute recordings are compared to test for one shared error mechanism versus several.
Tags
Description
We tested a timely scientific question: Does the brain (neocortex) fundamentally compute predictions (so-called predictive coding)? Specifically, do neurons signal every kind of surprise with *one* mechanism or *many*? We made use of open science (a free Allen Institute OpenScope dataset from a public archive).
README excerpt
# ai_oscp_neuro Analysis toolkit for the **[OpenScope Community Project](https://allenneuraldynamics.github.io/openscope-community-predictive-processing/)** dataset — Allen Institute for Neural Dynamics. > [!WARNING] > **Under development.** All results are preliminary and may change following further code audits. ## Background & motivation The [OpenScope Community Project on Predictive Processing](https://allenneuraldynamics.github.io/openscope-community-predictive-processing/) asks how the brain implements **predictive processing**. Predictive processing theorizes that the neocortex learns to predict upcoming sensory input and signals *prediction errors* when reality violates expectation. The main aim of this project is to derive whether different kinds of prediction violations (e.g., a sensory mismatch/oddball, a broken sensorimotor contingency, an omitted stimulus) are computed by **a variety of mechanisms and circuits** (H0) or by a **common canonical mechanism** that is repeated across the brain (H1). To decide between these two hypotheses, the same test battery of "mismatch" paradigms is recorded using **three distinct spatial scales in mouse visual cortex**: (1) Glutamate vesicle imaging (**SLAP2**), (2) Two-photon calcium imaging (**Mesoscope**), and (3) High density linear multiarray single neuron recordings across up to six brain areas simultaneously (**Neuropixels**). Using these three techniques in tandem, error signals can be compared from dendr