Skip to main content

Anokhin's Cognitome: Neural Hypernetwork and the Subject Problem

K. V. Anokhin proposes to describe the brain as a hypernetwork of neuronal groups with cognitive properties. The proposal asks which organization can count as the subject of successive experiences. An anatomical connection, a learned ensemble, a trial-level report and a self-model score are different data. Their relations require explicit measurement. Anokhin’s proposal

Distinguishing the subject from the measured response​

A fixed-point equation has no unique phenomenal interpretation. The identity channel fixes every state; a reset channel fixes any chosen target. Both admit a fixed state in the absence of a learned repertoire or an observer capable of a report. To support a correspondence with a subject, distinguish interventions on the candidate organization from those on an input, a decision or a reporting channel.

Testing a neural grouping​

Specify how a group is identified in the recording, what behaviour it predicts, and how it is tracked across trials. Compare a model retaining higher-order interactions with one retaining the same individual activity and pairwise statistics. A held-out change of context or stimulus can discriminate them. If they predict the same full records under every available intervention, those records cannot decide between the proposed implementations.

What a map between the models would have to preserve​

A map from a neural hypernetwork to Γ\Gamma needs a source category of admissible interventions, a target model and a test of composition. Predict the result of two interventions in succession using parameters fixed from single-intervention data. A mismatch can expose an insufficient encoder. Agreement supports the specified predictions; assignment of the same felt quality needs a further observation. Composition of observations · Phenomenal distinctions