Search

The basal ganglia are important for action initiation, selection, and motor learning. The input level, the striatum, receives input preferentially from the cortex and thalamus and is to 95% composed of striatal projection neurons (SPNs) with sparse GABAergic collaterals targeting distal dendrites of neighboring SPNs, in a distance-dependent manner. The remaining 5% are GABAergic and cholinergic interneurons. Our aim here is to investigate the role of surround inhibition for the intrinsic function of the striatum. Large-scale striatal networks of 20 to 40 thousand neurons were simulated with detailed multicompartmental models of different cell types, corresponding to the size of a module of the dorsolateral striatum, like the forelimb area (mouse). The effect of surround inhibition on dendritic computation and network activity was investigated, while groups of SPNs were activated. The SPN-induced surround inhibition in distal dendrites shunted effectively the corticostriatal EPSPs. The size of dendritic plateau-like potentials within the specific dendritic segment was both reduced and enhanced by inhibition, due to the hyperpolarized membrane potential of SPNs and the reversal-potential of GABA. On a population level, the competition between two subpopulations of SPNs was found to depend on the distance between the two units, the size of each unit, the activity level in each subgroup and the dopaminergic modulation of the dSPNs and iSPNs. The SPNs provided the dominating source of inhibition within the striatum, while the fast-spiking interneuron mainly had an initial effect due to short-term synaptic plasticity as shown in with ablation of the synaptic interaction., QC 20231121

Biophysically detailed multi-compartment models are powerful tools to explore computational principles of the brain and also serve as a theoretical framework to generate algorithms for artificial intelligence (AI) systems. However, the expensive computational cost severely limits the applications in both the neuroscience and AI fields. The major bottleneck during simulating detailed compartment models is the ability of a simulator to solve large systems of linear equations. Here, we present a novel Dendritic Hierarchical Scheduling (DHS) method to markedly accelerate such a process. We theoretically prove that the DHS implementation is computationally optimal and accurate. This GPU-based method performs with 2-3 orders of magnitude higher speed than that of the classic serial Hines method in the conventional CPU platform. We build a DeepDendrite framework, which integrates the DHS method and the GPU computing engine of the NEURON simulator and demonstrate applications of DeepDendrite in neuroscience tasks. We investigate how spatial patterns of spine inputs affect neuronal excitability in a detailed human pyramidal neuron model with 25,000 spines. Furthermore, we provide a brief discussion on the potential of DeepDendrite for AI, specifically highlighting its ability to enable the efficient training of biophysically detailed models in typical image classification tasks., QC 20231031

Purpose of the work is to research of the world science publications dynamics on the synchronization. Methods. The research methods are the statistical methods of data processing. Results. The emphasis in the study of synchronization over the past twenty years has shifted from physical and technical sciences to neuroscience with Asian countries domination., QC 20230630

The differences in the shape, form and location of neurons are closely linked to their function. Being able to accurately and efficiently reconstruct neurons digitally in a three-dimensional space is necessary for the acquisition of knowledge in this research field. Automation through software helps optimise efficiency, yet manual reconstructions are often preferred. This thesis therefore aims to help standardise the research field more and facilitate communication and collaborative efforts by evaluating three software, Vaa3D, Neutube and NCTracer, in regards to the reconstruction algorithms' accuracy, efficiency, consistency and user experience with the user interface in order to deduce their advantages and shortcomings. A downloadable and executable Java program, which compares similarities between two reconstructions, and scripts were written to measure these parameters. Vaa3D had higher accuracy and a significantly lower execution time, but Neutube and NCTracer showcased more stability and consistent results. Additionally, NCTracer proved to be more intuitive to use. All software exhibited their own drawbacks, but the information presented can aid in improving the software or the development of new software surpassing prior ones.

Quantization is a widely adopted technique for deep neural networks to reduce the memory and computational resources required. However, when quantized, most models would need a suitable calibration process to keep their performance intact, which requires data from the target domain, such as a fraction of the dataset used in model training and model validation (i.e. calibration dataset). In this study, we investigate the use of synthetic data as a substitute for the calibration with real data for the quantization method. We propose a data generation method based on Generative Adversarial Networks that are trained prior to the model quantization step. We compare the performance of models quantized using data generated by StyleGAN2-ADA and our pre-trained DiStyleGAN, with quantization using real data and an alternative data generation method based on fractal images. Overall, the results of our experiments demonstrate the potential of leveraging synthetic data for calibration during the quantization process. In our experiments, the percentage of accuracy degradation of the selected models was less than 0.6%, with our best performance achieved on MobileNetV2 (0.05%). The code is available at: https://github.com/ThanosM97/gsoc2022-openvino

Simulation of large-scale networks of neurons is an important approach to understanding and interpreting experimental data from healthy and diseased brains. Owing to the rapid development of simulation software and the accumulation of quantitative data of different neuronal types, it is possible to predict both computational and dynamical properties of local microcircuits in a ‘bottom-up’ manner. Simulated data from these models can be compared with experiments and ‘top-down’ modelling approaches, successively bridging the scales. Here we describe an open source pipeline, using the software Snudda, for predicting microcircuit connectivity and for setting up simulations using the NEURON simulation environment in a reproducible way. We also illustrate how to further ‘curate’ data on single neuron morphologies acquired from public databases. This model building pipeline was used to set up a first version of a full-scale cellular level model of mouse dorsal striatum. Model components from that work are here used to illustrate the different steps that are needed when modelling subcortical nuclei, such as the basal ganglia., QC 20220406

QC 20230613

Recent Neural Architecture Search (NAS) solutions have produced impressive results training super-networks and then deriving subnetworks, a.k.a. child models that outperform expert-crafted models from a pre-defined search space. Efficient and robust subnetworks can be selected for resource-constrained edge devices, allowing them to perform well in the wild. However, constructing super-networks for arbitrary architectures is still a challenge that often prevents the adoption of these approaches. To address this challenge, we present BootstrapNAS, a software framework for automatic generation of super-networks for NAS. BootstrapNAS takes a pre-trained model from a popular architecture, e.g., ResNet- 50, or from a valid custom design, and automatically creates a super-network out of it, then uses state-of-the-art NAS techniques to train the super-network, resulting in subnetworks that significantly outperform the given pre-trained model. We demonstrate the solution by generating super-networks from arbitrary model repositories and make available the resulting super-networks for reproducibility of the results., Comment: Accepted at AAAI2022 - Practical Deep Learning in the Wild

This paper presents a description of STC Ltd. systems submitted to the NIST 2021 Speaker Recognition Evaluation for both fixed and open training conditions. These systems consists of a number of diverse subsystems based on using deep neural networks as feature extractors. During the NIST 2021 SRE challenge we focused on the training of the state-of-the-art deep speaker embeddings extractors like ResNets and ECAPA networks by using additive angular margin based loss functions. Additionally, inspired by the recent success of the wav2vec 2.0 features in automatic speech recognition we explored the effectiveness of this approach for the speaker verification filed. According to our observation the fine-tuning of the pretrained large wav2vec 2.0 model provides our best performing systems for open track condition. Our experiments with wav2vec 2.0 based extractors for the fixed condition showed that unsupervised autoregressive pretraining with Contrastive Predictive Coding loss opens the door to training powerful transformer-based extractors from raw speech signals. For video modality we developed our best solution with RetinaFace face detector and deep ResNet face embeddings extractor trained on large face image datasets. The final results for primary systems were obtained by different configurations of subsystems fusion on the score level followed by score calibration.

We present a post-training weight pruning method for deep neural networks that achieves accuracy levels tolerable for the production setting and that is sufficiently fast to be run on commodity hardware such as desktop CPUs or edge devices. We propose a data-free extension of the approach for computer vision models based on automatically-generated synthetic fractal images. We obtain state-of-the-art results for data-free neural network pruning, with ~1.5% top@1 accuracy drop for a ResNet50 on ImageNet at 50% sparsity rate. When using real data, we are able to get a ResNet50 model on ImageNet with 65% sparsity rate in 8-bit precision in a post-training setting with a ~1% top@1 accuracy drop. We release the code as a part of the OpenVINO(TM) Post-Training Optimization tool.

Resolving the electronic structure of single biological molecules in their native state was among the primary motivations behind X-ray free-electron lasers. The ultra-short pulses they produce can outrun the atomic motion induced by radiation damage, but the electronic structure of the sample is still significantly modified from its original state. This paper explores the decoherence of the scattered signal induced by temporal evolution of the electronic structure in the sample molecule. It is shown that the undamaged electron density of a single-molecule sample can often be retrieved using only the two most occupied modes from the coherent mode decomposition of the partially coherent diffraction fluence.

The basal ganglia play an important role in decision making and selection of action primarily based on input from cortex, thalamus, and the dopamine system. Their main input structure, striatum, is central to this process. It consists of two types of projection neurons, together representing 95% of the neurons, and 5% of interneurons, among which are the cholinergic, fast-spiking, and low threshold-spiking subtypes. The membrane properties, somadendritic shape, and intrastriatal and extrastriatal synaptic interactions of these neurons are quite well described in the mouse, and therefore they can be simulated in sufficient detail to capture their intrinsic properties, as well as the connectivity. We focus on simulation at the striatal cellular/microcircuit level, in which the molecular/subcellular and systems levels meet. We present a nearly full-scale model of the mouse striatum using available data on synaptic connectivity, cellular morphology, and electrophysiological properties to create a microcircuit mimicking the real network. A striatal volume is populated with reconstructed neuronal morphologies with appropriate cell densities, and then we connect neurons together based on appositions between neurites as possible synapses and constrain them further with available connectivity data. Moreover, we simulate a subset of the striatum involving 10,000 neurons, with input from cortex, thalamus, and the dopamine system, as a proof of principle. Simulation at this biological scale should serve as an invaluable tool to understand the mode of operation of this complex structure. This platform will be updated with new data and expanded to simulate the entire striatum., QC 20200525

Resolving the electronic structure of single biological molecules in their native state was among the primary motivations behind X-ray free-electron lasers. The ultra-short pulses they produce can outrun the atomic motion induced by radiation damage, but the electronic structure of the sample is still significantly modified from its original state. This paper explores the decoherence of the scattered signal induced by temporal evolution of the electronic structure in the sample molecule. It is shown that the undamaged electron density of a single-molecule sample can often be retrieved using only the two most occupied modes from the coherent mode decomposition of the partially coherent diffraction fluence.

Resolving the electronic structure of single biological molecules in their native state was among the primary motivations behind X-ray free-electron lasers. The ultra-short pulses they produce can outrun the atomic motion induced by radiation damage, but the electronic structure of the sample is still significantly modified from its original state. This paper explores the decoherence of the scattered signal induced by temporal evolution of the electronic structure in the sample molecule. It is shown that the undamaged electron density of a single-molecule sample can often be retrieved using only the two most occupied modes from the coherent mode decomposition of the partially coherent diffraction fluence.

Resolving the electronic structure of single biological molecules in their native state was among the primary motivations behind X-ray free-electron lasers. The ultra-short pulses they produce can outrun the atomic motion induced by radiation damage, but the electronic structure of the sample is still significantly modified from its original state. This paper explores the decoherence of the scattered signal induced by temporal evolution of the electronic structure in the sample molecule. It is shown that the undamaged electron density of a single-molecule sample can often be retrieved using only the two most occupied modes from the coherent mode decomposition of the partially coherent diffraction fluence.

In this work we present a new framework for neural networks compression with fine-tuning, which we called Neural Network Compression Framework (NNCF). It leverages recent advances of various network compression methods and implements some of them, such as sparsity, quantization, and binarization. These methods allow getting more hardware-friendly models which can be efficiently run on general-purpose hardware computation units (CPU, GPU) or special Deep Learning accelerators. We show that the developed methods can be successfully applied to a wide range of models to accelerate the inference time while keeping the original accuracy. The framework can be used within the training samples, which are supplied with it, or as a standalone package that can be seamlessly integrated into the existing training code with minimal adaptations. Currently, a PyTorch version of NNCF is available as a part of OpenVINO Training Extensions at https://github.com/openvinotoolkit/nncf., Comment: 13 pages, 1 figure

Speaker recognition systems based on deep speaker embeddings have achieved significant performance in controlled conditions according to the results obtained for early NIST SRE (Speaker Recognition Evaluation) datasets. From the practical point of view, taking into account the increased interest in virtual assistants (such as Amazon Alexa, Google Home, AppleSiri, etc.), speaker verification on short utterances in uncontrolled noisy environment conditions is one of the most challenging and highly demanded tasks. This paper presents approaches aimed to achieve two goals: a) improve the quality of far-field speaker verification systems in the presence of environmental noise, reverberation and b) reduce the system qualitydegradation for short utterances. For these purposes, we considered deep neural network architectures based on TDNN (TimeDelay Neural Network) and ResNet (Residual Neural Network) blocks. We experimented with state-of-the-art embedding extractors and their training procedures. Obtained results confirm that ResNet architectures outperform the standard x-vector approach in terms of speaker verification quality for both long-duration and short-duration utterances. We also investigate the impact of speech activity detector, different scoring models, adaptation and score normalization techniques. The experimental results are presented for publicly available data and verification protocols for the VoxCeleb1, VoxCeleb2, and VOiCES datasets., Comment: Submitted to Odyssey 2020

Resolving the electronic structure of single biological molecules in their native state was among the primary motivations behind X-ray free-electron lasers. The ultra-short pulses they produce can outrun the atomic motion induced by radiation damage, but the electronic structure of the sample is still significantly modified from its original state. This paper explores the decoherence of the scattered signal induced by temporal evolution of the electronic structure in the sample molecule. It is shown that the undamaged electron density of a single-molecule sample can often be retrieved using only the two most occupied modes from the coherent mode decomposition of the partially coherent diffraction fluence.

© The Author(s) 2020. Published by Oxford University Press on behalf of Nucleic Acids Research. Bacterial single-stranded DNA-binding proteins (SSBs) bind single-stranded DNA and help to recruit heterologous proteins to their sites of action. SSBs perform these essential functions through a modular structural architecture: the N-terminal domain comprises a DNA binding/tetramerization element whereas the C-terminus forms an intrinsically disordered linker (IDL) capped by a protein-interacting SSB-Ct motif. Here we examine the activities of SSB-IDL fusion proteins in which fluorescent domains are inserted within the IDL of Escherichia coli SSB. The SSB-IDL fusions maintain DNA and protein binding activities in vitro, although cooperative DNA binding is impaired. In contrast, an SSB variant with a fluorescent protein attached directly to the C-terminus that is similar to fusions used in previous studies displayed dysfunctional protein interaction activity. The SSB-IDL fusions are readily visualized in single-molecule DNA replication reactions. Escherichia coli strains in which wildtype SSB is replaced by SSB-IDL fusions are viable and display normal growth rates and fitness. The SSB-IDL fusions form detectible SSB foci in cells with frequencies mirroring previously examined fluorescent DNA replication fusion proteins. Cells expressing SSB-IDL fusions are sensitized to some DNA damaging agents. The results highlight the utility of SSB-IDL fusions for biochemical and cellular studies of genome maintenance reactions.

In this work we present a new efficient approach to Human Action Recognition called Video Transformer Network (VTN). It leverages the latest advances in Computer Vision and Natural Language Processing and applies them to video understanding. The proposed method allows us to create lightweight CNN models that achieve high accuracy and real-time speed using just an RGB mono camera and general purpose CPU. Furthermore, we explain how to improve accuracy by distilling from multiple models with different modalities into a single model. We conduct a comparison with state-of-the-art methods and show that our approach performs on par with most of them on famous Action Recognition datasets. We benchmark the inference time of the models using the modern inference framework and argue that our approach compares favorably with other methods in terms of speed/accuracy trade-off, running at 56 FPS on CPU. The models and the training code are available., Comment: 8 pages, 3 figures

The basal ganglia are involved in the motivational and habitual control of motor and cognitive behaviors. Striatum, the largest basal ganglia input stage, integrates cortical and thalamic inputs in functionally segregated cortico-basal ganglia-thalamic loops, and in addition the basal ganglia output nuclei control targets in the brainstem. Striatal function depends on the balance between the direct pathway medium spiny neurons (D1-MSNs) that express D1 dopamine receptors and the indirect pathway MSNs that express D2 dopamine receptors. The striatal microstructure is also divided into striosomes and matrix compartments, based on the differential expression of several proteins. Dopaminergic afferents from the midbrain and local cholinergic interneurons play crucial roles for basal ganglia function, and striatal signaling via the striosomes in turn regulates the midbrain dopaminergic system directly and via the lateral habenula. Consequently, abnormal functions of the basal ganglia neuromodulatory system underlie many neurological and psychiatric disorders. Neuromodulation acts on multiple structural levels, ranging from the subcellular level to behavior, both in health and disease. For example, neuromodulation affects membrane excitability and controls synaptic plasticity and thus learning in the basal ganglia. However, it is not clear on what time scales these different effects are implemented. Phosphorylation of ion channels and the resulting membrane effects are typically studied over minutes while it has been shown that neuromodulation can affect behavior within a few hundred milliseconds. So how do these seemingly contradictory effects fit together? Here we first briefly review neuromodulation of the basal ganglia, with a focus on dopamine. We furthermore use biophysically detailed multi-compartmental models to integrate experimental data regarding dopaminergic effects on individual membrane conductances with the aim to explain the resulting cellular level dopami

In this work, we outline the set of problems, which any Object Detection CNN faces when its development comes to the deployment stage and propose methods to deal with such difficulties. We show that these practices allow one to get Object Detection network, which can recognize two classes: vehicles and pedestrians and achieves more than 60 frames per second inference speed on Core$^{TM}$ i5-6500 CPU. The proposed model is built on top of the popular Single Shot MultiBox Object Detection framework but with substantial improvements, which were inspired by the discovered problems. The network has just 1.96 GMAC complexity and less than 7 MB model size. It is publicly available as a part of Intel$\circledR$ OpenVINO$^{TM}$ Toolkit.

This paper presents the Speech Technology Center (STC) replay attack detection systems proposed for Automatic Speaker Verification Spoofing and Countermeasures Challenge 2017. In this study we focused on comparison of different spoofing detection approaches. These were GMM based methods, high level features extraction with simple classifier and deep learning frameworks. Experiments performed on the development and evaluation parts of the challenge dataset demonstrated stable efficiency of deep learning approaches in case of changing acoustic conditions. At the same time SVM classifier with high level features provided a substantial input in the efficiency of the resulting STC systems according to the fusion systems results., Comment: 11 pages, 3 figures, accepted for Specom 2017

The neural control of movements in vertebrates is based on a set of modules, like the central pattern generator networks (CPGs) in the spinal cord coordinating locomotion. Sensory feedback is not required for the CPGs to generate the appropriate motor pattern and neither a detailed control from higher brain centers. Reticulospinal neurons in the brainstem activate the locomotor network, and the same neurons also convey signals from higher brain regions, such as turning/steering commands from the optic tectum (superior colliculus). A tonic increase in the background excitatory drive of the reticulospinal neurons would be sufficient to produce coordinated locomotor activity. However, in both vertebrates and invertebrates, descending systems are in addition phasically modulated because of feedback from the ongoing CPG activity. We use the lamprey as a model for investigating the role of this phasic modulation of the reticulospinal activity, because the brainstem-spinal cord networks are known down to the cellular level in this phylogenetically oldest extant vertebrate. We describe how the phasic modulation of reticulospinal activity from the spinal CPG ensures reliable steering/turning commands without the need for a very precise timing of on-or offset, by using a biophysically detailed large-scale (19,600 model neurons and 646,800 synapses) computational model of the lamprey brainstem-spinal cord network. To verify that the simulated neural network can control body movements, including turning, the spinal activity is fed to a mechanical model of lamprey swimming. The simulations also predict that, in contrast to reticulospinal neurons, tectal steering/turning command neurons should have minimal frequency adaptive properties, which has been confirmed experimentally., QC 20140505

This study addresses mechanisms for the generation and selection of visual behaviors in anamniotes. To demonstrate the function of these mechanisms, we have constructed an experimental platform where a simulated animal swims around in a virtual environment containing visually detectable objects. The simulated animal moves as a result of simulated mechanical forces between the water and its body. The undulations of the body are generated by contraction of simulated muscles attached to realistic body components. Muscles are driven by simulated motoneurons within networks of central pattern generators. Reticulospinal neurons, which drive the spinal pattern generators, are in turn driven directly and indirectly by visuomotor centers in the brainstem. The neural networks representing visuomotor centers receive sensory input from a simplified retina. The model also includes major components of the basal ganglia, as these are hypothesized to be key components in behavior selection. We have hypothesized that sensorimotor transformation in tectum and pretectum transforms the place-coded retinal information into rate-coded turning commands in the reticulospinal neurons via a recruitment network mimicking the layered structure of tectal areas. Via engagement of the basal ganglia, the system proves to be capable of selecting among several possible responses, even if exposed to conflicting stimuli. The anatomically based structure of the control system makes it possible to disconnect different neural components, yielding concrete predictions of how animals with corresponding lesions would behave. The model confirms that the neural networks identified in the lamprey are capable of responding appropriately to simple, multiple, and conflicting stimuli., QC 20131106

We show an alternative way of representing a Bayesian belief network by sensitivities and probability distributions. This representation is equivalent to the traditional representation by conditional probabilities, but makes dependencies between nodes apparent and intuitively easy to understand. We also propose a QR matrix representation for the sensitivities and/or conditional probabilities which is more efficient, in both memory requirements and computational speed, than the traditional representation for computer-based implementations of probabilistic inference. We use sensitivities to show that for a certain class of binary networks, the computation time for approximate probabilistic inference with any positive upper bound on the error of the result is independent of the size of the network. Finally, as an alternative to traditional algorithms that use conditional probabilities, we describe an exact algorithm for probabilistic inference that uses the QR-representation for sensitivities and updates probability distributions of nodes in a network according to messages from the neighbors., Comment: Appears in Proceedings of the Eleventh Conference on Uncertainty in Artificial Intelligence (UAI1995)

We consider probabilistic inference in general hybrid networks, which include continuous and discrete variables in an arbitrary topology. We reexamine the question of variable discretization in a hybrid network aiming at minimizing the information loss induced by the discretization. We show that a nonuniform partition across all variables as opposed to uniform partition of each variable separately reduces the size of the data structures needed to represent a continuous function. We also provide a simple but efficient procedure for nonuniform partition. To represent a nonuniform discretization in the computer memory, we introduce a new data structure, which we call a Binary Split Partition (BSP) tree. We show that BSP trees can be an exponential factor smaller than the data structures in the standard uniform discretization in multiple dimensions and show how the BSP trees can be used in the standard join tree algorithm. We show that the accuracy of the inference process can be significantly improved by adjusting discretization with evidence. We construct an iterative anytime algorithm that gradually improves the quality of the discretization and the accuracy of the answer on a query. We provide empirical evidence that the algorithm converges., Comment: Appears in Proceedings of the Thirteenth Conference on Uncertainty in Artificial Intelligence (UAI1997)

Although probabilistic inference in a general Bayesian belief network is an NP-hard problem, computation time for inference can be reduced in most practical cases by exploiting domain knowledge and by making approximations in the knowledge representation. In this paper we introduce the property of similarity of states and a new method for approximate knowledge representation and inference which is based on this property. We define two or more states of a node to be similar when the ratio of their probabilities, the likelihood ratio, does not depend on the instantiations of the other nodes in the network. We show that the similarity of states exposes redundancies in the joint probability distribution which can be exploited to reduce the computation time of probabilistic inference in networks with multiple similar states, and that the computational complexity in the networks with exponentially many similar states might be polynomial. We demonstrate our ideas on the example of a BN2O network -- a two layer network often used in diagnostic problems -- by reducing it to a very close network with multiple similar states. We show that the answers to practical queries converge very fast to the answers obtained with the original network. The maximum error is as low as 5% for models that require only 10% of the computation time needed by the original BN2O model., Comment: Appears in Proceedings of the Twelfth Conference on Uncertainty in Artificial Intelligence (UAI1996)

Here, we investigate the role of sensory feedback in gait generation and transition by using a three-dimensional, neuro-musculo-mechanical model of a salamander with realistic physical parameters. Activation of limb and axial muscles were driven by neural output patterns obtained from a central pattern generator (CPG) which is composed of simulated spiking neurons with adaptation. The CPG consists of a body-CPG and four limb-CPGs that are interconnected via synapses both ipsilaterally and contralaterally. We use the model both with and without sensory modulation and four different combinations of ipsilateral and contralateral coupling between the limb-CPGs. We found that the proprioceptive sensory inputs are essential in obtaining a coordinated lateral sequence walking gait (walking). The sensory feedback includes the signals coming from the stretch receptor like intraspinal neurons located in the girdle regions and the limb stretch receptors residing in the hip and scapula regions of the salamander. On the other hand, walking trot gait (trotting) is more under central (CPG) influence compared to that of the peripheral or sensory feedback. We found that the gait transition from walking to trotting can be induced by increased activity of the descending drive coming from the mesencephalic locomotor region and is helped by the sensory inputs at the hip and scapula regions detecting the late stance phase. More neurophysiological experiments are required to identify the precise type of mechanoreceptors in the salamander and the neural mechanisms mediating the sensory modulation., QC 20111117, LAMPETRA

In this paper we introduce comparative analysis for different types of equalization schemes, based on the minimum mean square error (MMSE) optimization. The following types of equalizers were compared: linear equalization, decision feedback equalization (DFE) and turbo equalization. Performance and complexity of these schemes were tested for Single Carrier Frequency Division Multiple Access (SC-FDMA) system with Single Input Single Output (SISO) antenna configuration. SC-FDMA is a common technique, which is used in the UTRA LTE Uplink, so the results of complexity and performance analysis could be applied to find the appropriate equalization algorithm to be used in the Uplink channel of the LTE – the famous standard in 4G telecommunications. Simulation results in the end in this paper show bit error ratio (BER) and modulation error ratio (MER) for compared schemes.

In this paper we introduce comparative analysis for different types of equalization schemes, based on the minimum mean square error (MMSE) optimization. The following types of equalizers were compared: linear equalization, decision feedback equalization (DFE) and turbo equalization. Performance and complexity of these schemes were tested for Single Carrier Frequency Division Multiple Access (SC-FDMA) system with Single Input Single Output (SISO) antenna configuration. SC-FDMA is a common technique, which is used in the UTRA LTE Uplink, so the results of complexity and performance analysis could be applied to find the appropriate equalization algorithm to be used in the Uplink channel of the LTE – the famous standard in 4G telecommunications. Simulation results in the end in this paper show bit error ratio (BER) and modulation error ratio (MER) for compared schemes.

The vertebrate central nervous system is organized in modules that independently execute sophisticated tasks. Such modules are flexibly controlled and operate with a considerable degree of autonomy. One example is locomotion generated by spinal central pattern generator networks (CPGs) that shape the detailed motor output. The level of activity is controlled from brainstem locomotor command centers, which in turn, are under the control of the basal ganglia. By using a biophysically detailed, full-scale computational model of the lamprey CPG (10,000 neurons) and its brainstem/forebrain control, we demonstrate general control principles that can adapt the network to different demands. Forward or backward locomotion and steering can be flexibly controlled by local synaptic effects limited to only the very rostral part of the network. Variability in response properties within each neuronal population is an essential feature and assures a constant phase delay along the cord for different locomotor speeds., QC 20100525

The different neural control systems involved in goal-directed vertebrate locomotion are reviewed. They include not only the central pattern generator networks in the spinal cord that generate the basic locomotor synergy and the brainstem command systems for locomotion but also the control systems for steering and control of body orientation (posture) and finally the neural structures responsible for determining which motor programs should be turned on in a given instant. The role of the basal ganglia is considered in this context. The review summarizes the available information from a general vertebrate perspective, but specific examples are often derived from the lamprey, which provides the most detailed information when considering cellular and network perspectives., QC 20120120

The lamprey is one of the few vertebrates in which the neural control system for goal-directed locomotion including steering and control of body orientation is well described at a cellular level. In this report we review the modeling of the central pattern-generating network, which has been carried out based on detailed experimentation. In the same way the modeling of the control system for steering and control of body orientation is reviewed, including neuromechanical simulations and robotic devices., QC 20111107

Locomotor burst generation is simulated using a full-scale network model of the unilateral excitatory interneuronal population. Earlier small-scale models predicted that a population of excitatory neurons would be sufficient to produce burst activity, and this has recently been experimentally confirmed. Here we simulate the hemicord activity induced under various experimental conditions, including pharmacological activation by NMDA and AMPA as well as electrical stimulation. The model network comprises a realistic number of cells and synaptic connectivity patterns. Using similar distributions of cellular and synaptic parameters, as have been estimated experimentally, a large variation in dynamic characteristics like firing rates, burst, and cycle durations were seen in single cells. On the network level an overall rhythm was generated because the synaptic interactions cause partial synchronization within the population. This network rhythm not only emerged despite the distributed cellular parameters but relied on this variability, in particular, in reproducing variations of the activity during the cycle and showing recruitment in interneuronal populations. A slow rhythm (0.4-2 Hz) can be induced by tonic activation of NMDA-sensitive channels, which are voltage dependent and generate depolarizing plateaus. The rhythm emerges through a synchronization of bursts of the individual neurons. A fast rhythm (4-12 Hz), induced by AMPA, relies on spike synchronization within the population, and each burst is composed of single spikes produced by different neurons. The dynamic range of the fast rhythm is limited by the ability of the network to synchronize oscillations and depends on the strength of synaptic connections and the duration of the slow after hyperpolarization. The model network also produces prolonged bouts of rhythmic activity in response to brief electrical activations, as seen experimentally. The mutual excitation can sustain long-lasting activity for a realist, QC 20100525 QC 20111221

QC 20120301

A long standing belief in neuroscience has been that the brain and specifically the neocortex obtains its computational power by massive parallelism. Albeit conceptually appealing, this notion that effective processing requires large networks has not been possible to test in detailed simulations. In one project, we intend to study the generation of theta activity in the entorhinal-hippocampal system. Several simulation studies indicate that frequency and synchronization of the oscillation generated may depend on density of connectivity and/or geometry of connections. In a second project, we are studying how a model of early visual processing scales towards realistic sizes. To effectively evaluate the model, it must be scaled up to sizes where processing demands from the input given are sufficiently high, and where network size is made sufficiently large to process this information. We have in preliminary studies tested two parallel simulators. One is a version of pGENESIS supporting MPI from University of Sunderland, UK. The other is Split, a software produced in our own laboratory. Both have been tested on an Itanium2 cluster. Tests include variable number of processors and scaling number of neurons/compartments or number of synapses. In these simulations, average spike frequency in the network is also varied. The aim is to identify main bottle-necks. For instance, we foresee the need to parallelize the construction/layout of synapses., QC 20120113

The spinal CPG of the lamprey is modeled using a chain of nonlinear oscillators. Each oscillator represents a small neuron population capable of bursting under mixed NMDA and AMPA drive. Parameters of the oscillator are derived from detailed conductance-based neuron models. Analysis and simulations of dynamics of a single oscillator, a chain of locally coupled excitatory oscillators and a chain of two pairs of excitatory and inhibitory oscillators in each segment are done. The roles of asymmetric couplings and additional rostral drive for generation of a traveling wave with one cycle per chain length in a realistic frequency range are studied., QC 20100525 QC 20111221

Rhythmic locomotor-related activity can be induced in the isolated hemi-spinal cord of lamprey during bath application of D-glutamate or NMDA (Cangiano and Grillner, 2003). This bursting activity is not dependent on glycinergic inhibition but relies on mutual glutamatergic excitation among network interneurons. The possibility of such oscillatory activity was suggested by earlier simulations (Hellgren-Kotaleski et al. 1999). Here the underlying mechanisms are further examined using both detailed and reduced mathematical models. The detailed network model comprises a population of compartmental excitatory interneurones with Na+, K+, Ca2+, KCa channels as well as two Ca-pools. The synaptic interactions are mediated by AMPA receptors and voltage-dependent NMDA receptors, as established experimentally. This model reproduces the main experimental observations on both cell and network level, including the slow (NMDA/Mg2+ dependent) and the fast rhythm. Burst frequency can be modulated by changing the AMPA and/or NMDA drive, the latter providing only a narrow dynamic range. Further, the distributed network of the entire hemi-cord has been simulated. A weakly asymmetric rostro-caudal connectivity (stronger descending) could support a uniform intersegmental phase lag along most of the spinal cord, whereas a symmetric connectivity could not. The intersegmental phase lag is effectively controlled (forward and backward direction) by adding excitation or inhibition to the most rostral segments. The detailed model was progressively reduced until only the most important (slow) currents remained. The dynamics of the reduced model followed that of the detailed model. Ca influx and activation of KCa currents was shown to play a key role in the burst generation., QC 20111221

Straight locomotion in the lamprey is, at the segmental level, characterized by alternating bursts of motor activity with equal duration and spike frequency on the left and the right sides of the body. Lateral turns are characterized by three main changes in this pattern: (1) in the turn cycle, the spike frequency, burst duration, and burst proportion (burst duration/cycle duration) increase on the turning side; (2) the cycle duration increases in both the turn cycle and the succeeding cycle; and (3) in the cycle succeeding the turn cycle, the burst duration increases on the non-turning side (rebound). We investigated mechanisms for the generation of turns in single-segment models of the lamprey locomotor spinal network. Activation of crossing inhibitory neurons proved a sufficient mechanism to explain all three changes in the locomotor rhythm during a fictive turn. Increased activation of these cells inhibits the activity of the opposite side during the prolonged burst of the turn cycle, and slows down the locomotor rhythm. Secondly, this activation of the crossing inhibitory neurons is accompanied by an increased calcium influx into the cells. This gives a suppressed activity on the turning side and a contralateral rebound after the turn, through activation of calcium-dependent potassium channels., QC 20100525 QC 20111221

Consequences of synaptic plasticity in the lamprey spinal CPG are analyzed by means of simulations. This is motivated by the effects substance P (a tachykinin) and serotonin (5-hydroxytryptamin; 5-HT) have on synaptic transmission in the locomotor network. Activity-dependent synaptic depression and potentiation have recently been shown experimentally using paired intracellular recordings. Although normally activity-dependent plasticity presumably does not contribute to the patterning of network activity, this changes in the presence of the neuromodulators substance P and 5-HT, which evoke significant plasticity. Substance P can induce a faster and larger depression of inhibitory connections but potentiation of excitatory inputs, whereas 5-HT induces facilitation of both inhibitory and excitatory inputs. Changes in the amplitude of the first postsynaptic potential are also seen. These changes could thus be a potential mechanism underlying the modulatory role these substances have on the rhythmic network activity. The aim of the present study has been to implement the activity dependent synaptic depression and facilitation induced by substance P and 5-HT into two alternative models of the lamprey spinal locomotor network, one relying on reciprocal inhibition for bursting and one in which each hemicord is capable of oscillations. The consequences of the plasticity of inhibitory and excitatory connections are then explored on the network level. In the intact spinal cord, tachykinins and 5-HT, which can be endogenously released, increase and decrease the frequency of the alternating left-right burst pattern, respectively. The frequency decreasing effect of 5-HT has previously been explained based on its conductance decreasing effect on K underlying the postspike afterhyperpolarization (AHP). The present simulations show that short-term synaptic plasticity may have strong effects on frequency regulation in the lamprey spinal CPG. In the network model relying on reciprocal, QC 20100525 QC 20111221

NR 20140805

Consequences of synaptic plasticity in the lamprey spinal CPG are analyzed by means of simulations. This is motivated by the effects substance P (a tachykinin) and serotonin (5-hydroxytryptamin; 5-HT) have on synaptic transmission in the locomotor network. Activity-dependent synaptic depression and potentiation have recently been shown experimentally using paired intracellular recordings. Although normally activity-dependent plasticity presumably does not contribute to the patterning of network activity, this changes in the presence of the neuromodulators substance P and 5-HT, which evoke significant plasticity. Substance P can induce a faster and larger depression of inhibitory connections but potentiation of excitatory inputs, whereas 5-HT induces facilitation of both inhibitory and excitatory inputs. Changes in the amplitude of the first postsynaptic potential are also seen. These changes could thus be a potential mechanism underlying the modulatory role these substances have on the rhythmic network activity. The aim of the present study has been to implement the activity dependent synaptic depression and facilitation induced by substance P and 5-HT into two alternative models of the lamprey spinal locomotor network, one relying on reciprocal inhibition for bursting and one in which each hemicord is capable of oscillations. The consequences of the plasticity of inhibitory and excitatory connections are then explored on the network level. In the intact spinal cord, tachykinins and 5-HT, which can be endogenously released, increase and decrease the frequency of the alternating left-right burst pattern, respectively. The frequency decreasing effect of 5-HT has previously been explained based on its conductance decreasing effect on K(Ca) underlying the postspike afterhyperpolarization (AHP). The present simulations show that short-term synaptic plasticity may have strong effects on frequency regulation in the lamprey spinal CPG. In the network model relying on recipr, NR 20140805

Steering maneuvers in vertebrates are characterized by asymmetric modulation of the cycle duration and the intensity of the symmetric rhythmic locomotor activity. In the lamprey in vitro model system, turns can be evoked by electrical skin stimuli applied to one side of the head, which give rise to descending unilateral excitatory commands. Turns are observed as increased activity on one side of the spinal cord, followed by a rebound on the other. We investigated the generation of turns in single-segment models of the lamprey locomotor spinal network, and were able to reproduce all main experimental results. Sufficient mechanisms to explain changes in the locomotor rhythm, including rebound, are asymmetric activation of crossing inhibitory neurons, accompanied by a calcium influx in these neurons., NR 20140805

Consequences of synaptic plasticity in the lamprey spinal CPG are analyzed. This is motivated by the experimentally found effects substance P and 5-hydroxytryptamin (5-HT) have on the inhibitory and excitatory synaptic transmission. The effects can be a change of the amplitude of the postsynaptic potentials as well as induction of an activity-dependent facilitation or depression during repetitive activation. Simulations show that network level effects (i.e. swimming frequency) of substance P and 5-HT can to a substantial part be explained based on their effects on the plasticity of the synaptic transmission., NR 20140805

Consequences of synaptic plasticity in the lamprey spinal CPG are analyzed. This is motivated by the experimentally found effects substance P and 5-hydroxytryptamin (5-HT) have on the inhibitory and excitatory synaptic transmission. The effects can be a change of the amplitude of the postsynaptic potentials as well as induction of an activity-dependent facilitation or depression during repetitive activation. Simulations show that network level effects (i.e. swimming frequency) of substance P and 5-HT can to a substantial part be explained based on their effects on the plasticity of the synaptic transmission., QC 20100525 QC 20111221