Luis Cruz Cruz

Traveling waves of neuronal spiking activity are commonly observed across the brain, but their intrinsic function is still a matter of investigation. Experiments suggest that they may be valuable in the consolidation of memory or learning, indicating that consideration of traveling waves in the presence of plasticity might be important. A possible outcome of this consideration is that the synaptic pathways, necessary for the propagation of these waves, will be modified by the waves themselves. This will create a feedback loop where both the traveling waves and the strengths of the available synaptic pathways will change. To computationally investigate this, we model a sheet of cortical tissue by considering a quasi two-dimensional network of model neurons locally connected with plastic synaptic weights using Spike-Timing Dependent Plasticity (STDP). By using different stimulation conditions (central, stochastic, and alternating stimulation), we demonstrate that starting from a random network, traveling waves with STDP will form and strengthen propagation pathways. With progressive formation of traveling waves, we observe increases in synaptic weight along the direction of wave propagation, increases in propagation speed when pathways are strengthened over time, and an increase in the local order of synaptic weights. We also present evidence that the interaction between traveling waves and plasticity can serve as a mechanism of network-wide competition between available pathways. With an improved understanding of the interactions between traveling waves and synaptic plasticity, we can approach a fuller understanding of mechanisms of learning, computation, and processing within the brain.

Spiking neural networks (SNNs) are the next-generation neural networks composed of biologically plausible neurons that communicate through trains of spikes. By modifying the plastic parameters of SNNs, including weights and time delays, SNNs can be trained to perform various AI tasks, although in general not at the same level of performance as typical artificial neural networks (ANNs). One possible solution to improve the performance of SNNs is to consider plastic parameters other than just weights and time delays drawn from the inherent complexity of the neural system of the brain, which may help SNNs improve their information processing ability and achieve brainlike functions. Here, we propose reference spikes as a new type of plastic parameters in a supervised learning scheme in SNNs. A neuron receives reference spikes through synapses providing reference information independent of input to help during learning, whose number of spikes and timings are trainable by error backpropagation. Theoretically, reference spikes improve the temporal information processing of SNNs by modulating the integration of incoming spikes at a detailed level. Through comparative computational experiments, we demonstrate using supervised learning that reference spikes improve the memory capacity of SNNs to map input spike patterns to target output spike patterns and increase classification accuracy on the MNIST, Fashion-MNIST, and SHD data sets, where both input and target output are temporally encoded. Our results demonstrate that applying reference spikes improves the performance of SNNs by enhancing their temporal information processing ability.Spiking neural networks (SNNs) are the next-generation neural networks composed of biologically plausible neurons that communicate through trains of spikes. By modifying the plastic parameters of SNNs, including weights and time delays, SNNs can be trained to perform various AI tasks, although in general not at the same level of performance as typical artificial neural networks (ANNs). One possible solution to improve the performance of SNNs is to consider plastic parameters other than just weights and time delays drawn from the inherent complexity of the neural system of the brain, which may help SNNs improve their information processing ability and achieve brainlike functions. Here, we propose reference spikes as a new type of plastic parameters in a supervised learning scheme in SNNs. A neuron receives reference spikes through synapses providing reference information independent of input to help during learning, whose number of spikes and timings are trainable by error backpropagation. Theoretically, reference spikes improve the temporal information processing of SNNs by modulating the integration of incoming spikes at a detailed level. Through comparative computational experiments, we demonstrate using supervised learning that reference spikes improve the memory capacity of SNNs to map input spike patterns to target output spike patterns and increase classification accuracy on the MNIST, Fashion-MNIST, and SHD data sets, where both input and target output are temporally encoded. Our results demonstrate that applying reference spikes improves the performance of SNNs by enhancing their temporal information processing ability.

Traveling waves of neuronal activity in the cortex have been observed in vivo. These traveling waves have been correlated to various features of observed cortical dynamics, including spike timing variability and correlated fluctuations in neuron membrane potential. Although traveling waves are typically studied as either strictly one-dimensional or two-dimensional excitations, here we investigate the conditions for the existence of quasi-one-dimensional traveling waves that could be sustainable in parts of the brain containing cortical minicolumns. For that, we explore a quasi-one-dimensional network of heterogeneous neurons with a biologically influenced computational model of neuron dynamics and connectivity. We find that background stimulus reliably evokes traveling waves in networks with local connectivity between neurons. We also observe traveling waves in fully connected networks when a model for action potential propagation speed is incorporated. The biological properties of the neurons influence the generation and propagation of the traveling waves. Our quasi-one-dimensional model is not only useful for studying the basic properties of traveling waves in neuronal networks; it also provides a simplified representation of possible wave propagation in columnar or minicolumnar networks found in the cortex.

Molecular dynamics simulations are used to provide insights into the molecular mechanisms accounting for binding of amyloid fibrils to lipid bilayers and to study the effect of cholesterol in this process. We show that electrostatic interactions play an important role in fibril–bilayer binding and cholesterol modulates this interaction. In particular, the interaction between positive residues and lipid head groups becomes more favorable in the presence of cholesterol. Consistent with experiments, we find that cholesterol enhances fibril–membrane binding.

Neuronal connectivity at the cellular level in the cerebral cortex is far from random, with characteristics that point to a hierarchical design with intricately connected neuronal clusters. Here we investigate computationally the effects of varying neuronal cluster connectivity on network synchronization for two different spatial distributions of clusters: one where clusters are arranged in columns in a grid and the other where neurons from different clusters are spatially intermixed. We characterize each case by measuring the degree of neuronal spiking synchrony as a function of the number of connections per neuron and the degree of intercluster connectivity. We find that in both cases as the number of connections per neuron increases, there is an asynchronous to synchronous transition dependent only on intrinsic parameters of the biophysical model. We also observe in both cases that with very low intercluster connectivity clusters have independent firing dynamics yielding a low degree of synchrony. More importantly, we find that for a high number of connections per neuron but intermediate intercluster connectivity, the two spatial distributions of clusters differ in their response where the clusters in a grid have a higher degree of synchrony than the clusters that are intermixed.