A rapid and efficient learning rule for biological neural circuits

Abstract

The dominant view in neuroscience is that changes in synaptic weights underlie learning. It is unclear, however, how the brain is able to determine which synapses should change, and by how much. This uncertainty stands in sharp contrast to deep learning, where changes in weights are explicitly engineered to optimize performance. However, the main tool for doing that, backpropagation, is not biologically plausible, and networks trained with this rule tend to forget old tasks when learning new ones. Here we introduce the Dendritic Gated Network (DGN), a variant of the Gated Linear Network, which offers a biologically plausible alternative to backpropagation. DGNs combine dendritic "gating" (whereby interneurons target dendrites to shape neuronal response) with local learning rules to yield provably efficient performance. They are significantly more data efficient than conventional artificial networks and are highly resistant to forgetting, and we show that they perform well on a variety of tasks, in some cases better than backpropagation. The DGN bears similarities to the cerebellum, where there is evidence for shaping of Purkinje cell responses by interneurons. It also makes several experimental predictions, one of which we validate with in vivo cerebellar imaging of mice performing a motor task.

Look-up and look-down neurons in the mouse visual thalamus during freely moving exploration

“…Consistent with our results, eye movements, largely suppressed in head-fixed preparations, are strongly driven by changes in upward and downward facing postures in freely moving animals [55]. The richer modulations observed in freely moving conditions could also be employed to support coordinate transformation and spatial navigation during spontaneous exploration [19,56] or to learn new visuomotor contingencies by gating visual inputs according to behavioural context [57]. Finally, behavioural modulation could part of an encoding scheme that predicts incoming visual inputs based on self-motion [18].…”

Training a spiking neuronal network model of visual-motor cortex to play a virtual racket-ball game using reinforcement learning

“…The decreases in performance after learning could be due to overlapped representation of multiple ball trajectories in visual-motor space so that changes in the circuit required to learn about one ball trajectory may interfere with the visual-motor representation of the other ball trajectory. These issues have been observed in ANNs and are commonly termed "catastrophic forgetting" [48,[70][71][72][73][74][75]. We are still exploring several strategies to overcome forgetting in our SNNs.…”

Dendro-plexing single input spikes by multiple synaptic contacts enriches the computational capabilities of cortical neurons and reduces axonal wiring

“…Several phenomenological models have attempted to explain how multiple synaptic contacts between the pre-synaptic and post-synaptic neurons are formed (Fares and Stepanyants 2009), but very few studies have tried to tackle the question of how might they be beneficial from a computational perspective, but see (Sezener et al 2021;Camp, Mandivarapu, and Estrada 2020;Jones and Kording 2021;Zhang, Hu, and Liu 2020;Hiratani and Fukai 2018;Acharya et al 2021). It is typically thought that this redundancy overcomes the problem of unreliable synaptic vesicle release, which results in unreliable signal transmission between the pre-synaptic and the post-synaptic neurons (Rudolph et al 2015).…”

“…Zhang et al (Zhang, Hu, and Liu 2020) model multiple contacts in the context of deep artificial neural networks but demonstrate no tangible computational benefit. Several other studies use multiple synaptic contacts in the context of artificial neural networks, demonstrating some computational benefits, sometimes without explicitly addressing the use of multiple synaptic contacts (Camp, Mandivarapu, and Estrada 2020;Jones and Kording 2021;Sezener et al 2021).…”