See Also: Book Notes, (me), Notes on Consciousness, Consciousness and the Brain, Ancient Origins of Consciousness, Sapolsky: Behave, Consciousness: Confessions, How Emotions Are Made, The Quest for Consciousness, Happiness Hypothesis, Blank Slate, Info Viz & Perception, On Intelligence, Neuroscience of Human Relationships, Human: Makes Us Unique, Thinking, Fast and Slow, Righteous Mind, Ravenous Brain

Predictive Coding Animation AND Predictive Coding Conceptual Scematic

Gentle Introduction to Predictive Coding

Predictive Coding / Prediction / Prediction Error

The brain makes prediction about the world and our bodies and pushes those down the seconsory hierarchy. The sensory regions only pass informaton to the brain when it senses the prediction does not match the input. This becomes a Prediction Error (PE) and is the information sent up the hierarchy. If enough errors during sensory integration, then we become conscious of a novel situation. It depends on Generative models of reality - predictions based on prior experience using Bayesian math.

What is the right term?
"predictive coding" consciousness - About 40,800 results (0.47 seconds) And
"prediction processing" consciousness About 1,330 results (0.40 seconds)
"predictive processing" consciousness About 20,100 results (0.47 seconds) image ModelAtRest.png

7. Predictive Coding & Processing

Predictive Coding is a theory of how the brain sends codes from one part of the brain to another part of the brain. In essence, it posits that we use mental models of the world to predict what we are going to see and hear and then our perceptual systems confirm what we believe or not. The Cortical Columns are the 'prediction units'.

Here is a Gentle Introduction to Predictive Coding presented as a 'lay service' at First Parish.

There are four 'predictive processing theories': predictive coding, hierarchical temporal memory (HTM), bayesian inference, and Adaptive Resonance Theory (ART).

Model Error Axis
Model Error Axis

Visual and Motor Generative Models

Bidirectional Interaction between Visual and Motor Generative Models using Predictive Coding and Active Inference
Louis Annabi∗, Alexandre Pitti, Mathias Quoy
Are there really two models? Or one distrubuted. Model Error Axis

Amortized Prediction

Top Down AND Bottom Up precision signals

2021-08-03 I added the top down "precision expectations" to the schematic for now based on:

Predictive Coding: a Theoretical and Experimental Review

2021 - Beren Millidge, Anil Seth, Christopher L Buckley -

"There remains an intrinsic tension, however, between these two perspectives on precision in the literature. The first interprets precision as a bottom-up ‘objective’ measure of the intrinsic variance in the sensory data and then, deeper in the hierarchy, the intrinsic variance of activities at later processing stages. This contrasts strongly with views of precision as serving a general purpose adaptive modulatory function as in attention."

Good source with intro:

Primer on Predictive Processing

"The goal of this short chapter, aimed at philosophers, is to provide an overview and brief explanation of some central concepts involved in predictive processing (PP). Even those who consider themselves experts on the topic may find it helpful to see how the central terms are used in this collection. To keep things simple, we will first informally define a set of features important to predictive processing, supplemented by some short explanations and an alphabetic glossary."


Prediction and Allostasis - Big Picture

Paper: Intrinsic Functional Connectivity is Organized as Three Interdependent Gradients
Jiahe Zhang, Olamide Abiose, Yuta Katsumi, Alexandra Touroutoglou, Bradford C. Dickerson & Lisa Feldman Barrett
Scientific Reports volume 9, Article number: 15976 (2019)

This figure is a gold mine

This has the full bidirectional arc from model to world and back again with some very specific pathways.

Idea #1 create DTI streamers between the 105 seed points for a handful of representational percepts - the red Ferrari zooms by. Some notion of model data flowing and sense data flowing and prediction errors and precision signals flowing along the tracks.

ALSO! Check out WebGL 3D version of Figure 2.

See Below: Ouden 2012 Great Introduction

Predictive Processing: A Canonical Cortical Computation

Georg B. Keller and Thomas D. Mrsic-Flogel

Great diagrams! Evidence for Predictive Processing in Cortical Circuits Behavioral Evidence

With the discovery of the simple cells in cat primary vi-sual cortex (Hubel and Wiesel, 1959), the feature detector rapidly became the dominant narrative for our thinking about cortical function (Martin, 1994). This concept has been a guiding principle for scientific inquiry; it is apparent not only in the concept of receptive fields of neurons in visual cortex, but also in place cells (O'Keefe and Dostrovsky, 1971), grid cells (Hafting et al., 2005), face cells (Perrett et al., 1982), and concept cells (Quiroga et al., 2005). Once sensory systems of the brain have extracted an invariant representation from the sensory input, a separate part of the brain is then tasked with deciding and acting upon that representation. Following David Marr, we will call this the representational framework for describing the function of neocortex (Marr, 1982).

They concluded that one simple strategy to solve this problem of distinguishing self-generated sensory feedback from externally generated input in general would be to cancel the predictable consequences of sensory feedback using an efference copy of a motor command. This requires that the brain has a mechanism to transform the efference copy of the motor command into the sensory coordinate system to cancel the reafferent sensory feed-back. This transformed version of the efference copy is often referred to as a corollary discharge. Conceputally, such transfor-mations, or internal models, are equivalent to a simulation of the external world and function to make predictions of sensory input. Kenneth Craik formulated this idea in the early 1949s

Cortical circuits implement predictive processing. This requires at least three components: a comparator circuit that computes the prediction error between bottom-up input and predictions, a circuit to maintain an internal representation that gives rise to predictions, and a modulating or gating signal that sets the precision or weight of the prediction error. The circuit elements required to generate prediction errors are present in each module of the neocortex. Cortical areas receive bottom-up input from the thalamus or other cortical areas as well as extensive top-down inputs from many nearby and distal cortical areas and higher-order thalamic nuclei.

The error computation is carried out by two separate prediction-error circuits: one to signal more and one to signal less input than predicted (Figure 2). We will refer to these two types of prediction error as positive prediction error and negative prediction error.

We assume that perception is linked to the internal represen-tation of the world and that we only perceive a stimulus if the internal representation for that stimulus is active... First, an internal representation requires a circuit mechanism that maintains the activity in a population of neurons for the time a stimulus is perceived. A number of plausible mechanisms have been proposed for the persistence of neural activity, including strong and selective recurrent excitation between coactive neuronal assemblies within the cortex or via thalamo-cortical loops. Moreover, internal representations may not require stable patterns of activity, but could be maintained using dynamic attractors.

In either case, neurons representing the internal model are expected to exhibit more sustained and dense activity than neurons that function as comparators.

Second, internal representation neurons should make connections within the area they reside as well as provide top-down input to lower areas within the same sensory modality and/or project to asso-ciated cortical areas dedicated to other modalities. Finally, as internal representations need to be updated by prediction errors, the neurons encoding the internal representation should be densely connected with the comparator circuit encoding the same feature. Interestingly, these functional and anatomical characteristics are hallmarks of a subset of cortical neurons prevalent in deeper layers

Hybrid Predictive Coding: Inferring, Fast and Slow

Hybrid Predictive Coding: Inferring, Fast and Slow
Alexander Tschantz, Beren Millidge, Anil K Seth, Christopher L Buckley

Eplains amortized prediction as the vocabulary idea from <"./hawkinsi.html">On Intelligence

"Here, we propose that the feedforward sweep can be understood as performing amortized inference and recurrent processing can be understood as performing iterative inference. We propose a hybrid predictive coding network that combines both iterative and amortized inference in a principled manner by describing both in terms of a dual optimization of a single objective function. We show that the resulting scheme can be implemented in a biologically plausible neural architecture that approximates Bayesian inference utilising local Hebbian update rules."

Review paper: How prediction errors shape perception, attention, and motivation

Hanneke E. M. den Ouden, Peter Kok and Floris P . de Lange

Predictive coding posits the existence of separate “prediction” units (P) and prediction error units (PE) within each cortical column, the basic cortical computational module. There are both intrinsic (within the cortical column) and extrinsic (between columns) connections between P and PE units.
Prediction and Prediction Error signals travel to and from specific layers.
Large PEs lead to updating of predictions, which are then sent forward as input to higher cortical areas (via superficial layers, L2/3), as well as backward to update predictions in lower areas (via deep layers, L5/6). In this scheme, the excitatory feedback from higher to lower selection function of the basal ganglia is not limited to action selection or reinforcement learning. Given that the basal ganglia receive cortical, limbic, and brainstem inputs and thus have access to motivational,affective,cognitive,and motor information, they may form the final common pathway where information from a wide variety of sources is integrated and then guide selection among cortical representations, actions and goals. I
Functional differences then arise from differences in the input sources and output targets. For example, outcome predictions in dorsolateral stria-tum will be derived from inputs from sensorimotor areas and will lead to action selection, whereas nucleus accumbens, with inputs from the amygdala and VTA, will report reward PEs in a rein-forcement learning task and allow motivational aspects to guide behavior. Hippocampal and prefrontal glutamatergic inputs to the ventral striatum may provide the top-down context information or predictions to modulate neural activity.

Canonical microcircuits for predictive coding , (2012) by Bastos, Andre M, Usrey, W Martin, Adams, Rick A, Mangun, George R, Fries, Pascal and Friston, Karl J [bib] Neurons & Layers Prediction Flow

Specifically, valid prior expectations allow for selection of the proper hypotheses (i.e., activating relevant P units) in advance of stimulation, facilitation of perception (Bar, 2004)


—- deviate - attracter states

This paper talks about the body model in layer 4 used to generate predictions. The Body Model Theory of Somatosensory Cortex

"Sensory Gating" ... "Most importantly, is sensory gating under top-down or bottom-up control?". In the Predictive Coding Model, The sensory gating would is the model being pushed down to the senses. Cortical modulation of sensory flow during active touch in the rat whisker system with a lay person explanation here: Sensory Perception Is Not a One-Way Street

2022-07-31 <> started 2018.08.26 <> jch