Fundamental Computational Unit Research Articles

Towards Interpretable Physical‐Conceptual Catchment‐Scale Hydrological Modeling Using the Mass‐Conserving‐Perceptron

AbstractWe investigate the applicability of machine learning technologies to the development of parsimonious, interpretable, catchment‐scale hydrologic models using directed‐graph architectures based on the mass‐conserving perceptron (MCP) as the fundamental computational unit. Here, we focus on architectural complexity (depth) at a single location, rather than universal applicability (breadth) across large samples of catchments. The goal is to discover a minimal representation (numbers of cell‐states and flow paths) that represents the dominant processes that can explain the input‐state‐output behaviors of a given catchment, with particular emphasis given to simulating the full range (high, medium, and low) of flow dynamics. We find that a “HyMod Like” architecture with three cell‐states and two major flow pathways achieves such a representation at our study location, but that the additional incorporation of an input‐bypass mechanism significantly improves the timing and shape of the hydrograph, while the inclusion of bi‐directional groundwater mass exchanges significantly enhances the simulation of baseflow. Overall, our results demonstrate the importance of using multiple diagnostic metrics for model evaluation, while highlighting the need for properly selecting and designing the training metrics based on information‐theoretic foundations that are better suited to extracting information across the full range of flow dynamics. This study sets the stage for interpretable regional‐scale MCP‐based hydrological modeling (using large sample data) by using neural architecture search to determine appropriate minimal representations for catchments in different hydroclimatic regimes.

Distinct Alterations in Dendritic Spine Morphology in the Absence of β-Neurexins.

Dendritic spines are essential for synaptic function because they constitute the postsynaptic compartment of the neurons that receives the most excitatory input. The extracellularly shorter variant of the presynaptic cell adhesion molecules neurexins, β-neurexin, has been implicated in various aspects of synaptic function, including neurotransmitter release. However, its role in developing or stabilizing dendritic spines as fundamental computational units of excitatory synapses has remained unclear. Here, we show through morphological analysis that the deletion of β-neurexins in hippocampal neurons in vitro and in hippocampal tissue in vivo affects presynaptic dense-core vesicles, as hypothesized earlier, and, unexpectedly, alters the postsynaptic spine structure. Specifically, we observed that the absence of β-neurexins led to an increase in filopodial-like protrusions in vitro and more mature mushroom-type spines in the CA1 region of adult knockout mice. In addition, the deletion of β-neurexins caused alterations in the spine head dimension and an increase in spines with perforations of their postsynaptic density but no changes in the overall number of spines or synapses. Our results indicate that presynaptic β-neurexins play a role across the synaptic cleft, possibly by aligning with postsynaptic binding partners and glutamate receptors via transsynaptic columns.

Large-scale single-neuron speech sound encoding across the depth of human cortex

Understanding the neural basis of speech perception requires that we study the human brain both at the scale of the fundamental computational unit of neurons and in their organization across the depth of cortex. Here we used high-density Neuropixels arrays1–3 to record from 685 neurons across cortical layers at nine sites in a high-level auditory region that is critical for speech, the superior temporal gyrus4,5, while participants listened to spoken sentences. Single neurons encoded a wide range of speech sound cues, including features of consonants and vowels, relative vocal pitch, onsets, amplitude envelope and sequence statistics. Neurons at each cross-laminar recording exhibited dominant tuning to a primary speech feature while also containing a substantial proportion of neurons that encoded other features contributing to heterogeneous selectivity. Spatially, neurons at similar cortical depths tended to encode similar speech features. Activity across all cortical layers was predictive of high-frequency field potentials (electrocorticography), providing a neuronal origin for macroelectrode recordings from the cortical surface. Together, these results establish single-neuron tuning across the cortical laminae as an important dimension of speech encoding in human superior temporal gyrus.

Design and analysis of hybrid 10T adder for low power applications

Recently, a growing focus has been on the need for ultra-low-voltage (ULV) functioning to reduce energy usage. The full adder is a fundamental computational arithmetic unit in various image and signal processing applications. This study presents a novel architectural design for a 1-bit adder that utilizes 10 transistors. The concept relies on an energy-efficient hybrid logic multiplexer approach incorporating Gate Diffusion Input (GDI) logic. The primary goal of the suggested design for a hybrid full adder is to decrease power consumption while concurrently decreasing the required area. The suggested innovative architectural design for a full adder incorporates level restoration carry logic and ensures a complete swing output voltage. The designs have been analyzed in a standard environment using 45-nm CMOS process technology, with various supply voltages utilized. The evaluation of the adder cells focuses on speed, power consumption, power-delay-product (PDP), and space efficiency compared to the existing full adders. The recommended full adder cells were subjected to comprehensive simulation tests utilizing the Mentor Graphics environment. The results suggest that these cells surpass their counterparts, exhibiting a minimum of 80 % enhancements in their power-delay product (PDP) parameters. Also the comparisons have been made between the proposed full adders with and without level restoration techniques. The proposed level restoration technique based full adder yields better performance in terms of power and delay when compared to proposed full adder without level restoration.

Cryogenic CMOS Circuits and Systems: Challenges and Opportunities in Designing the Electronic Interface for Quantum Processors

Quantum computing could potentially offer faster solutions for some of today's classically intractable problems using quantum processors as computational support for quantum algorithms [1]. Quantum processors, in the most frequent embodiment, comprise an array of quantum bits (qubits), the fundamental computational unit of a quantum computer. Unlike conventional bits, qubits can take a coherent state ranging from |0g to |1g on a continuous sphere, known as the Bloch sphere (Figure 1). When the state of the qubit, represented by a vector on the Bloch sphere, is on the equator of such a sphere, qubits are said to be in maximum superposition. Entanglement is the second important quantum mechanical property of qubit states, where knowing the state of one qubit implies knowing the state of the other one as well.

Emerging ideas and tools to study the emergent properties of the cortical neural circuits for voluntary motor control in non-human primates.

For years, neurophysiological studies of the cerebral cortical mechanisms of voluntary motor control were limited to single-electrode recordings of the activity of one or a few neurons at a time. This approach was supported by the widely accepted belief that single neurons were the fundamental computational units of the brain (the “neuron doctrine”). Experiments were guided by motor-control models that proposed that the motor system attempted to plan and control specific parameters of a desired action, such as the direction, speed or causal forces of a reaching movement in specific coordinate frameworks, and that assumed that the controlled parameters would be expressed in the task-related activity of single neurons. The advent of chronically implanted multi-electrode arrays about 20 years ago permitted the simultaneous recording of the activity of many neurons. This greatly enhanced the ability to study neural control mechanisms at the population level. It has also shifted the focus of the analysis of neural activity from quantifying single-neuron correlates with different movement parameters to probing the structure of multi-neuron activity patterns to identify the emergent computational properties of cortical neural circuits. In particular, recent advances in “dimension reduction” algorithms have attempted to identify specific covariance patterns in multi-neuron activity which are presumed to reflect the underlying computational processes by which neural circuits convert the intention to perform a particular movement into the required causal descending motor commands. These analyses have led to many new perspectives and insights on how cortical motor circuits covertly plan and prepare to initiate a movement without causing muscle contractions, transition from preparation to overt execution of the desired movement, generate muscle-centered motor output commands, and learn new motor skills. Progress is also being made to import optical-imaging and optogenetic toolboxes from rodents to non-human primates to overcome some technical limitations of multi-electrode recording technology.

Rubrocerebellar Feedback Loop Isolates the Interposed Nucleus as an Independent Processor of Corollary Discharge Information in Mice.

Understanding cerebellar contributions to motor coordination requires deeper insight into how the output structures of the cerebellum, the cerebellar nuclei, integrate their inputs and influence downstream motor pathways. The magnocellular red nucleus (RNm), a brainstem premotor structure, is a major target of the interposed nucleus (IN), and has also been described in previous studies to send feedback collaterals to the cerebellum. Because such a pathway is in a key position to provide motor efferent information to the cerebellum, satisfying predictions about the use of corollary discharge in cerebellar computations, we studied it in mice of both sexes. Using anterograde viral tracing, we show that innervation of cerebellum by rubrospinal neuron collaterals is remarkably selective for the IN compared with the cerebellar cortex. Optogenetic activation of the pathway in acute mouse brain slices drove IN activity despite small amplitude synaptic currents, suggesting an active role in IN information processing. Monosynaptic transsynaptic rabies tracing indicated the pathway contacts multiple cell types within the IN. By contrast, IN inputs to the RNm targeted a region that lacked inhibitory neurons. Optogenetic drive of IN inputs to the RNm revealed strong, direct excitation but no inhibition of RNm neurons. Together, these data indicate that the cerebellar nuclei are under afferent control independent of the cerebellar cortex, potentially diversifying its roles in motor control.SIGNIFICANCE STATEMENT The common assumption that all cerebellar mossy fibers uniformly collateralize to the cerebellar nuclei and cortex underlies classic models of convergent Purkinje influence on cerebellar output. Specifically, mossy fibers are thought to both directly excite nuclear neurons and drive polysynaptic feedforward inhibition via Purkinje neurons, setting up a fundamental computational unit. Here we present data that challenge this rule. A dedicated cerebellar nuclear afferent comprised of feedback collaterals from premotor rubrospinal neurons can directly modulate IN output independent of Purkinje cell modulation. In contrast to the IN-RNm pathway, the RNm-IN feedback pathway targets multiple cell types, potentially influencing both motor output pathways and nucleo-olivary feedback.

Life-Long Genetic and Functional Access to Neural Circuits Using Self-Inactivating Rabies Virus

SummaryNeural networks are emerging as the fundamental computational unit of the brain and it is becoming progressively clearer that network dysfunction is at the core of a number of psychiatric and neurodegenerative disorders. Yet, our ability to target specific networks for functional or genetic manipulations remains limited. Monosynaptically restricted rabies virus facilitates the anatomical investigation of neural circuits. However, the inherent cytotoxicity of the rabies largely prevents its implementation in long-term functional studies and the genetic manipulation of neural networks. To overcome this limitation, we developed a self-inactivating ΔG-rabies virus (SiR) that transcriptionally disappears from the infected neurons while leaving permanent genetic access to the traced network. SiR provides a virtually unlimited temporal window for the study of network dynamics and for the genetic and functional manipulation of neural circuits in vivo without adverse effects on neuronal physiology and circuit function.

Function approximation by hardware spiking neural network

Spiking neural networks (SNN) represent a special class of artificial neural networks, where neu-ron models communicate by sequences of spikes. SNNs are often referred to as the third generation of neural networks that highly inspired from natural computing in the brain and recent advances in neuroscience. In this paper we implement biologically-inspired, hardware-realizable SNN architecture using integrate-and-fire units, which is capable of approximating a real-valued function. Based on the results of MATLAB simulations, hardware synthesis and FPGA implementation, it is demonstrated that the implemented hardware can approximate linear and nonlinear functions with low minimum relative error. This framework may represent a fundamental computational unit for the development of artificial SNN, opening new perspectives in pattern recognition tasks.

Multiplier-less digital implementation of neuron–astrocyte signalling on FPGA

Experimental observations have shown that astrocytes are active partner in neural information processing which are mainly carried out by neurons. In this research, we investigate this issue and provide a computationally efficient digital circuit for neuron–astrocyte interactions. The firing dynamics of the neuron is described by Izhikevich model (spiking and bursting activities) and the calcium dynamics of a single astrocyte explained by a proposed model adapted from a functional model introduced by Postnov and colleagues. The new linear model makes it possible to design an efficient multiplier-less hardware architecture for digital implementation of neuron–astrocyte interaction on field-programmable gate array (FPGA). Using the proposed neuron–astrocyte circuit and based on the results of MATLAB simulation, hardware synthesis and FPGA implementation, it is demonstrated that the proposed digital astrocyte is able to change the neuron spiking frequency through bidirectional signaling. This circuit may represent a fundamental computational unit for the development of artificial neuron–astrocyte networks, opening new perspectives in pattern recognition tasks.

A novel low-energy CNFET-based full adder cell using pass-transistor logic

Full adder is the most critical and fundamental computational unit within processors. A new single-bit full adder cell is presented in this paper in order to cell efficiency. The new design is based on pass-transistor logic PTL and carbon nanotube field effect transistor CNFET technology. This combination resolves the problem of voltage drop in PTL. The proposed structure is an excellent trade-off between speed and power efficiency factors. Simulations are made by Synopsys HSPICE with 32 nm CNFET in various realistic conditions. The results demonstrate that the proposed adder cell outperforms other state-of-the-art designs in terms of energy consumption, and it is an appropriate candidate to be used in large circuits.

A model of order-selectivity based on dynamic changes in the balance of excitation and inhibition produced by short-term synaptic plasticity

Determining the order of sensory events separated by a few hundred milliseconds is critical to many forms of sensory processing, including vocalization and speech discrimination. Although many experimental studies have recorded from auditory order-sensitive and order-selective neurons, the underlying mechanisms are poorly understood. Here we demonstrate that universal properties of cortical synapses-short-term synaptic plasticity of excitatory and inhibitory synapses-are well suited for the generation of order-selective neural responses. Using computational models of canonical disynaptic circuits, we show that the dynamic changes in the balance of excitation and inhibition imposed by short-term plasticity lead to the generation of order-selective responses. Parametric analyses predict that among the forms of short-term plasticity expressed at excitatory-to-excitatory, excitatory-to-inhibitory, and inhibitory-to-excitatory synapses, the single most important contributor to order-selectivity is the paired-pulse depression of inhibitory postsynaptic potentials (IPSPs). A topographic model of the auditory cortex that incorporates short-term plasticity accounts for both context-dependent suppression and enhancement in response to paired tones. Together these results provide a framework to account for an important computational problem based on ubiquitous synaptic properties that did not yet have a clearly established computational function. Additionally, these studies suggest that disynaptic circuits represent a fundamental computational unit that is capable of processing both spatial and temporal information.

Assessing Degree of Functionality of Simple and Microcolumnar Neuronal Networks by Measuring Aspects of Information Processing: A Computational Approach

Some areas in the cerebral cortex are characterized by microcolumns: arrays of interconnected neurons which may constitute a fundamental computational unit in the brain. These microcolumns (also known as minicolumns), are formed by small, vertical columns of neurons that can span all layers in the gray matter. Although correlative studies have established that microcolumns can lose their primary characteristics in normal aging, neurological, and neurodegenerative diseases, the exact function of these structures has not been established. Using computer simulations of highly detailed neuronal networks, we study whether there is a functional advantage in neuronal networks with a microcolumnar geometry as compared to other geometrically distributed networks. In particular, these simulations take into account microcolumnar, crystalline, and random distributions of neurons. At the assigned location of each neuron, we position unique, geometrically precise neurons and determine synaptic connectivity of the networks via three models: fully-connected, gaussian, and hypergeometric connectivity models, taking into account the physical distances between the neurons. By utilizing the neuronal simulation package NEURON, we perform functional tests on these neuronal networks, such as information maintenance and scaling effects of network growth. Results on the advantages and disadvantages of each network topology are presented, and arguments are formulated that could explain microcolumnar neuronal networks as a natural evolution to maximize information processing in the brain.

A neuron–astrocyte transistor-like model for neuromorphic dressed neurons

Experimental evidences on the role of the synaptic glia as an active partner together with the bold synapse in neuronal signaling and dynamics of neural tissue strongly suggest to investigate on a more realistic neuron-glia model for better understanding human brain processing. Among the glial cells, the astrocytes play a crucial role in the tripartite synapsis, i.e. the dressed neuron. A well-known two-way astrocyte-neuron interaction can be found in the literature, completely revising the purely supportive role for the glia. The aim of this study is to provide a computationally efficient model for neuron-glia interaction. The neuron-glia interactions were simulated by implementing the Li-Rinzel model for an astrocyte and the Izhikevich model for a neuron. Assuming the dressed neuron dynamics similar to the nonlinear input-output characteristics of a bipolar junction transistor, we derived our computationally efficient model. This model may represent the fundamental computational unit for the development of real-time artificial neuron-glia networks opening new perspectives in pattern recognition systems and in brain neurophysiology.

High-field fMRI unveils orientation columns in humans

Functional (f)MRI has revolutionized the field of human brain research. fMRI can noninvasively map the spatial architecture of brain function via localized increases in blood flow after sensory or cognitive stimulation. Recent advances in fMRI have led to enhanced sensitivity and spatial accuracy of the measured signals, indicating the possibility of detecting small neuronal ensembles that constitute fundamental computational units in the brain, such as cortical columns. Orientation columns in visual cortex are perhaps the best known example of such a functional organization in the brain. They cannot be discerned via anatomical characteristics, as with ocular dominance columns. Instead, the elucidation of their organization requires functional imaging methods. However, because of insufficient sensitivity, spatial accuracy, and image resolution of the available mapping techniques, thus far, they have not been detected in humans. Here, we demonstrate, by using high-field (7-T) fMRI, the existence and spatial features of orientation- selective columns in humans. Striking similarities were found with the known spatial features of these columns in monkeys. In addition, we found that a larger number of orientation columns are devoted to processing orientations around 90 degrees (vertical stimuli with horizontal motion), whereas relatively similar fMRI signal changes were observed across any given active column. With the current proliferation of high-field MRI systems and constant evolution of fMRI techniques, this study heralds the exciting prospect of exploring unmapped and/or unknown columnar level functional organizations in the human brain.

A statistically based density map method for identification and quantification of regional differences in microcolumnarity in the monkey brain

We present a statistical density map method derived from condensed matter physics to quantify microcolumns, the fundamental computational unit of the cerebral cortex. This method provides measures for microcolumnar strength, width, spacing, length, and periodicity. We applied this method to Nissl-stained 30 μm thick frozen sections from areas 46, TE, and TL of rhesus monkey brains, areas that differ visually in microcolumnarity and are associated with different cognitive functions. Our results indicate that microcolumns in these areas are similar in width, spacing, and periodicity, but are stronger (possess a higher neuronal density) in area TE, as compared to areas TL and 46. We modeled the effect of section orientation on microcolumnar spacing and demonstrated that this method provides an adequate estimate of spacing. We also modeled disruption of microcolumnarity by performing simulations that randomly displace neurons and demonstrated that displacements of only one neuronal diameter effectively eliminate microcolumnar organization. These results indicate that our density map method is sensitive enough to detect and quantify subtle differences in microcolumnar organization that may occur in the context of development, aging, and neuropathology, as well as between areas and species.

Age-related reduction in microcolumnar structure in area 46 of the rhesus monkey correlates with behavioral decline.

Many age-related declines in cognitive function are attributed to the prefrontal cortex, area 46 being especially critical. Yet in normal aging, studies indicate that neurons are not lost in area 46, suggesting that impairments result from more subtle processes. One cortical feature that is functionally important, but that has not been examined in normal aging because of a lack of efficient quantitative methods, is the vertical arrangement of neurons into microcolumns, a fundamental computational unit of the cortex. By using a density-map method derived from condensed-matter physics, we quantified microcolumns in area 46 of seven young and seven aged rhesus monkeys that had been cognitively tested. This analysis demonstrated that there is no age-related reduction in total neuronal density or in microcolumn width, length, or periodicity. There was, however, a statistically significant decrease in the strength of microcolumns, indicating microcolumnar disorganization. This reduction in strength was significantly correlated with age-related cognitive decline on tests of spatial working memory and recognition memory independent of the effect of age. Modeling demonstrated that random neuron displacements of approximately 30% of a neuronal diameter (<3 mum) produced the observed reduction in strength. Hence, it is possible that, with changes in dendrites and myelinated axons, subtle displacements of neurons occur that alter microcolumnar structure and correlate with age-induced dysfunction. Therefore, quantitative measurement of microcolumnar structure may provide a sensitive morphological method to assay microcolumnar function in aging and other conditions.

A hillslope-based hydrological model using catchment area and width functions

A flow-interval hillslope discretization scheme is proposed for catchment hydrological modelling. By this scheme, a two-dimensional catchment is simplified into a one-dimensional cascade of flow intervals linked by the main stream. Each flow interval comprises a set of parallel hillslopes. The hillslope is the fundamental computational unit in the hydrological model providing lateral inflow to the main stream. The size of hillslope is determined by the catchment area and width functions. Catchment runoff is the total of hillslope responses through the river routing. Tests in four Japanese catchments showed that the model performed well on simulating the overall water balance, general flow pattern, and daily and hourly hydrographs of a whole catchment, as well as simultaneous simulation in different subcatchments. Characteristics of catchment hydrological responses and model applicability are discussed.

Comparison of different distributed hydrological models for characterization of catchment spatial variability

Most physically-based hydrological models rely on either spatially distributed or lumped characterizations of topography and other spatial variations. Three disturbed hydrological models with different treatments of topography, namely the MIKE SHE, TOPMODEL and GB model (geomorphology-based hydrological model) are discussed in this paper. A regular square grid system is used by MIKE SHE for representation of catchment spatial variability and as the fundamental computation units. TOPMODEL characterizes the topography using a distributed topographic index, ln(a/tan β). The areas with the same ln(a/tan β) are assumed to be hydrologically similar. Then, the catchment is divided into a number of segments corresponding to the topographic index intervals. The GB model employs the catchment area function and width function to lump the topography and divide the catchment into a series of flow interval-hillslopes. The catchment spatial variations are averaged over each flow interval and represented by one-dimensional distribution functions with respect to the flow distance from the catchment outlet. The hillslope elements are the fundamental computational units in the GB model. The above three models are applied to the Seki River in Japan. The model structure and performance are compared and discussed in the paper. Copyright © 2000 John Wiley & Sons, Ltd.

A model of neocortex

Prompted by considerations about (i) the compositionality of cognitive functions, (ii) the physiology of individual cortical neurons, (iii) the role of accurately timed spike patterns in cortex, and (iv) the regulation of global cortical activity, we suggest that the dynamics of cortex on the 1-ms time scale may be described as the activation of circuits of the synfire-chain type (Abeles 1982, 1991). We suggest that the fundamental computational unit in cortex may be a wave-like spatio-temporal pattern of synfire type, and that the binding mechanism underlying compositionality in cognition may be the accurate synchronization of synfire waves that propagate simultaneously on distinct, weakly coupled, synfire chains. We propose that Hebbian synaptic plasticity may result in a superposition of synfire chains in cortical connectivity, whereby a given neuron participates in many distinct chains. We investigate the behaviour of a much-simplified model of cortical dynamics devised along these principles. Calculations and numerical experiments are performed based on an assumption of randomness of stored chains, in the style of statistical physics. It is demonstrated that: (i) there exists a critical value for the total length of stored chains; (ii) this storage capacity is linear in the network's size; (iii) the behaviour of the network around the critical point is characterized by the self-regulation of the number of synfire waves coactive in the network at any given time.