![]() Lastly, an overview of parameters and toolbox functions is given in Tables 1 and 2 at the end of the document.įig 1. 5), we present selected examples where intrinsic timescales play an important role. While of general interest, this section is not required for a general understanding of the toolbox. ![]() ![]() 3) before we derive the MR estimator and discuss technical details such as the impact of short trials (Sec. We then briefly focus on the neuroscience context (including a real-life example, Sec. In the following, we discuss how to apply the toolbox using a code example (Sec. Lastly, the toolbox calculates confidence intervals by default, when a trial structure is provided. It supports trial structures and we demonstrate how multiple trials can be combined to compensate for short individual trials. The main advantage of using our toolbox over a custom implementation to determine intrinsic timescales is that it provides a consistent way that can now be adopted across studies. in epidemiology or social sciences such as the timescale of epidemic spreading (from subsampled infection counts) or the timescale of opinion spreading (from subsampled social networks). Since our method is based on spreading processes in complex systems, it is applicable beyond neuroscience, e.g. Estimator” that implements MR to estimate the intrinsic timescale of spiking activity, even for heavily subsampled systems. Because the intrinsic timescale inferred by MR is invariant to spatial subsampling, one can infer it even when recording only a small set of units. However, we recently showed that this spatial subsampling only biases the magnitude of the autocorrelation function (of autoregressive processes) and that-despite the bias-the associated intrinsic timescale can still be inferred by using multi-step regression (MR). This subsampling problem is especially problematic in neuroscience, where even the most advanced electrode measurements can record at most a few thousand out of the billions of neurons in the brain. In experiments we approach this level: we typically sample only a small part of the system, sometimes only a single or a dozen of units. However, approaching the single-unit level, the magnitude of the autocorrelation function can be much smaller than expected, and can be disguised by noise. Such a spreading process typically features an exponentially decaying autocorrelation function, and the associated time constant is in principle accessible from the activity of each unit. One can consider spiking activity in a recurrent network as a branching or spreading process, where each presynaptic spike triggers on average a certain number m of postsynaptic spikes. The single neuron basically serves as a readout for the local network activity. Īlthough autocorrelations and the intrinsic timescale can be derived from single neuron activity, they characterize the dynamics within the whole recurrent network. More importantly, such decaying autocorrelations are also found in the network-spiking-dynamics recorded in the brain: Here, the intrinsic timescale serves as a measure to quantify working memory and unravels a temporal hierarchy of processing in primates. ), where the intrinsic timescale can be related to information storage and transfer. Exponentially decaying correlations are commonly found in recurrent networks (see e.g. Intuitively, the intrinsic timescale characterizes the decay time of an exponentially decaying autocorrelation function (in this work and in many contexts it is synonymous to the autocorrelation time). Recent discoveries in the field of computational neuroscience suggest a major role of the so-called intrinsic timescale for functional brain dynamics. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.Ĭompeting interests: The authors have declared that no competing interests exist. All authors acknowledge funding by the Max Planck Society. JZ is supported by the Joachim Herz Stiftung. įunding: FPS and JD were funded by the Volkswagen Foundation through the SMARTSTART Joint Training Program Computational Neuroscience. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.ĭata Availability: Referenced scripts are available at Simulation data are available at. ![]() Received: OctoAccepted: MaPublished: April 29, 2021Ĭopyright: © 2021 Spitzner et al. Estimator, a toolbox to determine intrinsic timescales from subsampled spiking activity. Citation: Spitzner FP, Dehning J, Wilting J, Hagemann A, P.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |