Encoding Time in Feedforward Trajectories of a Recurrent Neural Network Model

Hardy, Nicholas F.; Buonomano, Dean V.

doi:10.1162/neco_a_01041

Cited by 52 publications

(43 citation statements)

References 68 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We wondered how the variance in our network compared to Weber's law. According to Weber's law, the standard deviation of reactions in a timing task grows linearly with time [Gibbon, 1977, Hardy andBuonomano, 2018]. Since our model operates on a time scale that is behaviourally relevant, it is interesting to look at how the variability increases with increasing time.…”

Section: Properties Of the Modelmentioning

confidence: 99%

Learning spatiotemporal signals using a recurrent spiking network that discretizes time

Maes

Barahona

Clopath

2019

Preprint

View full text Add to dashboard Cite

Learning to produce spatiotemporal sequences is a common task the brain has to solve. The same neural substrate may be used by the brain to produce different sequential behaviours. The way the brain learns and encodes such tasks remains unknown as current computational models do not typically use realistic biologicallyplausible learning. Here, we propose a model where a spiking recurrent network of excitatory and inhibitory biophysical neurons drives a read-out layer: the dynamics of the recurrent network is constrained to encode time while the read-out neurons encode space. Space is then linked with time through plastic synapses that follow common Hebbian learning rules. We demonstrate that the model is able to learn spatiotemporal dynamics on a timescale that is behaviourally relevant. Learned sequences are robustly replayed during a regime of spontaneous activity. Author summaryThe brain has the ability to learn flexible behaviours on a wide range of time scales. Previous studies have successfully build spiking network models that learn a variety of computational tasks. However, often the learning involved is not local. Here, we investigate a model using biological-plausible plasticity rules for a specific computational task: spatiotemporal sequence learning. The architecture separates time and space into two different parts and this allows learning to bind space to time. Importantly, the time component is encoded into a recurrent network which exhibits sequential dynamics on a behavioural time scale. This network is then used as an engine to drive spatial read-out neurons. We demonstrate that the model can learn complicated spatiotemporal spiking dynamics, such as the song of a bird, and replay the song robustly.

show abstract

Section: Properties Of the Modelmentioning

confidence: 99%

Learning spatiotemporal signals using a recurrent spiking network that discretizes time

Maes

Barahona

Clopath

2019

Preprint

View full text Add to dashboard Cite

show abstract

“…with the time-encoding features of the neural trajectories during the SCT. Consequently, to test this idea we first characterized the properties of neuronal moving bumps [10,12,26] during this task. With this information we carried out simulations to determine whether the key features of the moving bumps were linked to the observed changes in curvature radius and variability as a function of duration in the neural state trajectories.…”

Section: Neural Population Trajectories and Evolving Activation Patternsmentioning

confidence: 99%

The amplitude in periodic neural state trajectories underlies the tempo of rhythmic tapping

Gámez¹,

Mendoza²,

Prado³

et al. 2018

Preprint

View full text Add to dashboard Cite

Our motor commands can be exquisitely timed according to the demands of the environment, and the ability to generate rhythms of different tempos is a hallmark of musical cognition. Yet, the neuronal basis behind rhythmic tapping remains elusive. Here we found that the activity of hundreds of primate MPC neurons show a strong periodic pattern that becomes evident when their activity is projected into a lower dimensional state space. We show that different tempos are encoded by circular trajectories that travelled at a constant speed but with different radii, and that this neuronal code is highly resilient to the number of participating neurons. Crucially, the changes in the amplitude of the oscillatory dynamics in neuronal state space are a signature of beat-based timing, regardless of whether it is guided by an external metronome or is internally controlled and is not the result of repetitive motor commands. Furthermore, the increase in amplitude and variability of the neural trajectories accounted for the scalar property of interval timing. In addition, we found that the interval-dependent increments in the radius of periodic neural trajectories are the result of larger number of neurons engaged in the production of longer intervals. Our results support the notion that beat-based timing during rhythmic behaviors is encoded in the radial curvature of periodic MPC neural population trajectories.

show abstract

“…The two most well-established explanations of human time perception are the internal clock model (2)(3)(4) and the population clock approach (5,6). In the internal clock approach, it is proposed that some regular physiological or neural process produces rhythmic ticks like the hands of a clock.…”

Section: Introductionmentioning

confidence: 99%

Trial-by-trial predictions of subjective time from human brain activity

Sherman

Fountas

Seth

et al. 2020

Preprint

View full text Add to dashboard Cite

Author ContributionsWR conceived of the study. MTS and WR designed and pre-registered the experiments and analyses. MTS collected, analyzed, and constructed models of human behavioral and neuroimaging data. ZF constructed the artificial network model and analyzed the data. MTS and WR wrote the manuscript. AKS and ZF provided critical revisions on the manuscript. AbstractHuman experience of time exhibits systematic, context-dependent deviations from objective clock time. For example, time is experienced differently at work than on holiday. The cognitive and neural bases of how time perception interacts with the content of experience remain unclear, and leading explanations of human time perception are not equipped to explain this interaction. We propose an alternative account of human time perception, based on the dynamics of sensory processing. Our approach naturally links content of experience with time perception through a common foundation in basic sensory processing. We provide evidence for this proposal in model-based analyses of the dynamics of perceptual processing in an artificial neural network and in the activity of human sensory cortex. Healthy human participants watched naturalistic, silent videos and estimated their duration while fMRI was acquired. The same videos were used as stimuli for the artificial network.We constructed a computational model that predicted video durations from salient events in the activity of the artificial network, or in participants' visual cortex. The artificial network model reproduced human-like duration estimates, including biases in estimation depending on the content of a given video. Most importantly, the model based on human visual cortex activity reproduced trial-by-trial biases in our human participants' subjective reports, whereas control models trained on auditory or somatosensory activity did not. Together, our results reveal that human subjective time is based on information arising during the processing of our dynamic sensory environment, providing a computational basis for an end-to-end account of time perception. Significance StatementOur perception of time depends on the contents of experience, reflected in expressions such as "a watched pot never boils". Prevailing accounts of human time perception cannot provide good explanations for this. We tested a new explanation: time perception arises from the processing of our dynamic sensory environment. Supporting this theory, human participants' duration reports of silent videos correlated with estimates we reconstructed from activity in the visual cortex of their brain while they watched those videos. This was not a brain-wide phenomenon; reconstructions based on other brain regions did not correlate with subjective reports of duration. Our results validate our new theory, linking content of experience and perception of time through a common foundation in basic sensory processing.

show abstract

Encoding Time in Feedforward Trajectories of a Recurrent Neural Network Model

Cited by 52 publications

References 68 publications

Learning spatiotemporal signals using a recurrent spiking network that discretizes time

Learning spatiotemporal signals using a recurrent spiking network that discretizes time

The amplitude in periodic neural state trajectories underlies the tempo of rhythmic tapping

Trial-by-trial predictions of subjective time from human brain activity

Contact Info

Product

Resources

About