Deep Predictive Learning in Neocortex and Pulvinar

Abstract

How do humans learn from raw sensory experience? Throughout life, but most obviously in infancy, we learn without explicit instruction. We propose a detailed biological mechanism for the widely embraced idea that learning is driven by the differences between predictions and actual outcomes (i.e., predictive error-driven learning). Specifically, numerous weak projections into the pulvinar nucleus of the thalamus generate top–down predictions, and sparse driver inputs from lower areas supply the actual outcome, originating in Layer 5 intrinsic bursting neurons. Thus, the outcome representation is only briefly activated, roughly every 100 msec (i.e., 10 Hz, alpha), resulting in a temporal difference error signal, which drives local synaptic changes throughout the neocortex. This results in a biologically plausible form of error backpropagation learning. We implemented these mechanisms in a large-scale model of the visual system and found that the simulated inferotemporal pathway learns to systematically categorize 3-D objects according to invariant shape properties, based solely on predictive learning from raw visual inputs. These categories match human judgments on the same stimuli and are consistent with neural representations in inferotemporal cortex in primates.

Gated recurrence enables simple and accurate sequence prediction in stochastic, changing, and structured environments

“…First, it relies only on the observation sequence (no supervision or reinforcement), leveraging prediction error signals, which have been found in the brain in many studies (Den Ouden et al, 2012;Eshel et al, 2013;Maheu et al, 2019). Importantly, in predictive coding (Rao & Ballard, 1999), the computation of prediction errors is part of the prediction process; here we are suggesting that it may also be part of the training process (as argued in (O'Reilly et al, 2021)). Second, relatively few iterations of training suffice (Fig.…”

Correcting the Hebbian Mistake: Toward a Fully Error-Driven Hippocampus

“…Although this form of error-driven learning might make sense computationally, how could something like this delta error signal emerge naturally from the hippocampal biology? First, as in our prior model of learning in the monosynaptic pathway (Ketz et al, 2013), we adopt the idea that the delta arises as a temporal difference between two states of activity over the CA3, which is also consistent with a broader understanding of how error-driven learning works in the neocortex; (O'Reilly, 1996;O'Reilly & Munakata, 2000;O'Reilly, Russin, Zolfaghar, & Rohrlich, 2021). The appropriate temporal difference over CA3 should actually emerge naturally from the additional delay associated with the propagation of the MF signal through the DG to the CA3, compared to the more direct PP signal from ECin to CA3.…”

Gated recurrence enables simple and accurate sequence prediction in stochastic, changing, and structured environments

“…Importantly, in predictive coding 65 , the computation of prediction errors is part of the prediction process; here we are suggesting that it may also be part of the training process (as argued in Ref. 119 ). The second aspect is that relatively few iterations of training suffice (Supplementary Fig.…”