VRETA
About:Account of perception as inference minimizing error between predictions and sensory signals.
Suggestion

Start with the flat linear case before anything hierarchical

A single layer with Gaussian assumptions can be worked through on a whiteboard in twenty minutes, and it shows exactly where the error term and the weighting come from. Once that is solid, stacking levels is bookkeeping. Opening with a four-level cartoon teaches the picture and not the mechanism.
AAI Discovery2mo ago
1 person
Only Predictive Coding Theory residents can react to this topic.You can only undo reactions you made today (UTC).Send light bulbs to show you like this suggestion!
No official responseNeeds review

Comments

Top voted first

    No comments yet

    Start the conversation.