Predictive Coding as a Neuromorphic Alternative to Backpropagation: A Critical Evaluation
This work addresses the problem of finding biologically plausible learning algorithms for neuromorphic systems, but it is incremental as it builds on existing PC variants and highlights limitations rather than proposing a new solution.
The paper critically evaluates predictive coding (PC) as a neuromorphic alternative to backpropagation, finding that current PC variants have time complexity at least as high as backpropagation and may have limited potential as a direct replacement.
Backpropagation has rapidly become the workhorse credit assignment algorithm for modern deep learning methods. Recently, modified forms of predictive coding (PC), an algorithm with origins in computational neuroscience, have been shown to result in approximately or exactly equal parameter updates to those under backpropagation. Due to this connection, it has been suggested that PC can act as an alternative to backpropagation with desirable properties that may facilitate implementation in neuromorphic systems. Here, we explore these claims using the different contemporary PC variants proposed in the literature. We obtain time complexity bounds for these PC variants which we show are lower-bounded by backpropagation. We also present key properties of these variants that have implications for neurobiological plausibility and their interpretations, particularly from the perspective of standard PC as a variational Bayes algorithm for latent probabilistic models. Our findings shed new light on the connection between the two learning frameworks and suggest that, in its current forms, PC may have more limited potential as a direct replacement of backpropagation than previously envisioned.