Flexible timing by temporal scaling of cortical responses

By Jing Wang, Devika Narain, Eghbal A. Hosseini, Mehrdad Jazayeri. Nature Neuroscience, 2017

407 citations recorded by the source corpus.

Musicians can perform at different tempos, speakers can control the cadence of their speech, and children can flexibly vary their temporal expectations of events. To understand the neural basis of such flexibility, we recorded from the medial frontal cortex of nonhuman primates trained to produce different time intervals with different effectors. Neural responses were heterogeneous, nonlinear, and complex, and they exhibited a remarkable form of temporal invariance: firing rate profiles were temporally scaled to match the produced intervals. Recording from downstream neurons in the caudate and from thalamic neurons projecting to the medial frontal cortex indicated that this phenomenon originates within cortical networks. Recurrent neural network models trained to perform the task revealed that temporal scaling emerges from nonlinearities in the network and that the degree of scaling is controlled by the strength of external input. These findings demonstrate a simple and general mechanism for conferring temporal flexibility upon sensorimotor and cognitive functions. Humans can deliberately control the timing of their actions but the neural mechanisms underlying such control are largely unknown. In this article, Wang, Narain and their colleagues report that such flexibility emerges in rhesus monkeys from the ability of their brain to flexibly control the speed at which cortical responses unfold in time.

Subfields: Cognitive Neuroscience

View Paper: https://www.semanticscholar.org/paper/1a4459d8d1def42611a88e13cfb692f9aac1eb89

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Cognitive Neuroscience

Citations in the corpus

Flexible sensorimotor computations through rapid reconfiguration of cortical dynamics (2018, unknown): “More recently, it was found that actions are initiated when the collective activity of neurons with both ramping and more complex activity patterns reach an action-triggering state (Churchland et al., 2006; Wang et al., 2018) and that flexible control of initiation time can be understood in terms of”
Linking Connectivity, Dynamics, and Computations in Low-Rank Recurrent Neural Networks. (2017, unknown): “This approach therefore directly links the connectivity in the network to the emerging low-dimensional dynamics, and shows that the dominant dimensions of activity are determined by a combination of feed-forward inputs and connectivity (Wang et al., 2018).”
The Neural Basis of Timing: Distributed Mechanisms for Diverse Functions (2018, background): “…affecting the BG including Parkinson’s disease, Huntington’s disease (Freeman et al., 1996), Tourette’s (Vicario et al., 2010), substance abuse (Wittmann et al., 2007), and attention deficit disorder (Noreika et al., 2013) have been associated with altered sensation of time or temporally…”
The remarkable robustness of surrogate gradient learning for instilling complex function in spiking neural networks (2020, background): “Similarly, the activity of artificial recurrent neural networks optimized to solve cognitive tasks resembles cortical activity in prefrontal [12, 13], medial frontal [14], and motor areas [15, 16] and inspire comparison and lively discussion.”
Brain-Score: Which Artificial Neural Network for Object Recognition is most Brain-Like? (2018, background): “One could also use entirely new types of comparison, such as precise temporal dynamics of neural responses that are ignored here, even though they are likely to play an important role in brain function (Wang et al., 2018), or causal manipulations that may constrain models more strongly (Rajalingham ”
Interpreting neural computations by examining intrinsic and embedding dimensionality of neural activity (2021, background): “Similarly, interactions between large populations of neurons in the frontal and parietal cortex are thought to support simple decision-making tasks that involve just a few relevant variables [3–5].”
Motor primitives in space and time via targeted gain modulation in cortical networks (2018, result): “In line with recent results [22, 23], we investigate if gain changes are able to control the speed of an intended movement (see Fig.”
Complementary roles of dimensionality and population structure in neural computations (2020, background): “(10) represents the component of the activity on the recurrent space [Wang et al., 2018; Remington et al., 2018] defined as the sub-space spanned by the output connectivity patterns {m}r=1.”
Computation Through Neural Population Dynamics (2020, background): “While commendable progress has been made toward characterizing inputs (e.g., Wang et al. 2018), how activity from one brain region affects another (e.g., Semedo et al. 2019), or how population covariance structure changes across tasks (e.g.,Gallego et al. 2018), further headway will likely require…”
The Encoding of Speech Sounds in the Superior Temporal Gyrus (2019, background): “Recurrent models have also provided useful insights into other domains related to cortical processing, demonstrating that contextual dependence is a fundamental part of neuronal function during complex cognitive behaviors (Mante et al., 2013; Sussillo and Abbott, 2009; Wang et al., 2018).”
Artificial Neural Networks for Neuroscientists: A Primer (2020, background): “While backpropagation in an MLP propagates gradient information from the final layer back (Equation 13), computing the gradient for an RNN involves propagating information backward in time (backpropagation-through-time, or BPTT) (Werbos, 1990).”