Figure 13:

Comparison between: a) The proposed thalamocortical temporal-difference predictive learning model (from Figure 2), versus b) The Bayesian-style explicit error (EE) coding model (Rao & Ballard, 1999; Friston, 2010, Bastos et al., 2012). The EE model holds that superficial (S, lamina 2/3) error-coding neurons receive the prediction via a net inhibitory top-down projection from higher-level deep layer (D) neurons, and an excitatory bottom-up projection representing the outcome, such that their activation represents the difference. To encode both signs of the error (omissions, false alarms) with positive-only spike rates, two separate populations of EE neurons would be required, or a more complicated deviation from tonic firing level scheme. Unambiguous evidence of such EE coding neurons has not been found (Walsh et al, 2020). In contrast, error signals in our proposed framework remain as a temporal difference between the two states of prediction vs. outcome, which enables all connectivity between cortical areas to be excitatory and always represent a positive encoding of either the prediction or outcome. In contrast, under EE, after one error subtraction at the lowest level, only error signals are hypothesized to flow forward to higher layers, meaning that the representations at higher layers are about increasingly higher-order errors, not positive encodings of the environmental state at increasing levels of abstraction. These are indicated by ? because they are difficult to picture intuitively, and they are inconsistent with extensive available data showing similar positive representations of the external world at all levels in the visual hierarchy. Although some frameworks make claims about temporal dynamics, these are not strongly constrained by the basic computational framework, so that also remains a question.