Top-down feedback in deep neural networks leads to functional differences during audiovisual integration
Artificial neural networks (ANNs) are an important tool for studying neural computation, but many features of the brain are not captured by standard ANN architectures. One notable missing feature in most ANN models is top-down feedback, that is projections from higher-order layers to lower-order layers in the network. Top-down feedback is ubiquitous in the brain, and it has a unique modulatory…
A recent study reveals the significance of top-down feedback in deep neural networks (ANNs) and its impact on audiovisual integration. The research team created a deep neural network model that replicates the essential functional characteristics of top-down feedback found in the neocortex of the human brain. This enabled the development of hierarchical recurrent ANN models that closely resemble the brain's architecture.
By examining various hierarchical recurrent architectures in an audiovisual integration task, the researchers discovered that hierarchical configurations resembling the human brain gave ANN models a slight visual bias, akin to human performance. This visual bias did not hinder the ANN models' overall performance on the task. Furthermore, the study indicates that different configurations of top-down feedback can make otherwise identical ANN models functionally distinct, setting them apart from traditional feedforward and laterally recurrent models.
This research underscores the importance of modulatory top-down feedback in biological brains and highlights how incorporating it into ANNs can influence their behavior and limit the solutions they are likely to discover.
Written by urgent.news from eLife's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
