scholarly journals Abstract representations emerge naturally in neural networks trained to perform multiple tasks

2021 ◽  
Author(s):  
W. Jeffrey Johnston ◽  
Stefano Fusi
Keyword(s):  
Neural Networks ◽  
Natural World ◽  
Brain Regions ◽  
Neural Population ◽  
Task Structure ◽  
Cognitive Variables ◽  
Population Activity ◽  
Data Set ◽  
Abstract Representations ◽  
Neurophysiological Studies

Humans and other animals demonstrate a remarkable ability to generalize knowledge across distinct contexts and objects during natural behavior. We posit that this ability depends on the geometry of the neural population representations of these objects and contexts. Specifically, abstract, or disentangled, neural representations -- in which neural population activity is a linear function of the variables important for making a decision -- are known to allow for this kind of generalization. Further, recent neurophysiological studies have shown that the brain has sufficiently abstract representations of some sensory and cognitive variables to enable generalization across distinct contexts. However, it is unknown how these abstract representations emerge. Here, using feedforward neural networks, we demonstrate a simple mechanism by which these abstract representations can be produced: The learning of multiple distinct classification tasks. We demonstrate that, despite heterogeneity in the task structure, abstract representations that enable reliable generalization can be produced from a variety of different inputs -- including standard nonlinearly mixed inputs, inputs that mimic putative representations from early sensory areas, and even simple image inputs from a standard machine learning data set. Thus, we conclude that abstract representations of sensory and cognitive variables emerge from the multiple behaviors that animals exhibit in the natural world, and may be pervasive in high-level brain regions. We make several specific predictions about which variables will be represented abstractly as well as show how these representations can be detected.

scholarly journals Continuous Attractor Neural Networks: Candidate of a Canonical Model for Neural Information Representation

F1000Research ◽  
2016 ◽  
Vol 5 ◽  
pp. 156 ◽  
Author(s):  
Si Wu ◽  
K Y Michael Wong ◽  
C C Alan Fung ◽  
Yuanyuan Mi ◽  
Wenhao Zhang
Keyword(s):  
Neural Networks ◽  
Dimensional Space ◽  
Spatial Location ◽  
Canonical Model ◽  
Neural Population ◽  
Information Representation ◽  
Population Activity ◽  
Neural Information ◽  
Complex Features ◽  
Low Dimensional

Owing to its many computationally desirable properties, the model of continuous attractor neural networks (CANNs) has been successfully applied to describe the encoding of simple continuous features in neural systems, such as orientation, moving direction, head direction, and spatial location of objects. Recent experimental and computational studies revealed that complex features of external inputs may also be encoded by low-dimensional CANNs embedded in the high-dimensional space of neural population activity. The new experimental data also confirmed the existence of the M-shaped correlation between neuronal responses, which is a correlation structure associated with the unique dynamics of CANNs. This body of evidence, which is reviewed in this report, suggests that CANNs may serve as a canonical model for neural information representation.

scholarly journals Coexistence of fast and slow gamma oscillations in one population of inhibitory spiking neurons

2019 ◽  
Author(s):  
Hongjie Bi ◽  
Marco Segneri ◽  
Matteo di Volo ◽  
Alessandro Torcini
Keyword(s):  
Gamma Oscillations ◽  
Brain Regions ◽  
Neural Population ◽  
Neural Firing ◽  
Phase Coupling ◽  
Population Activity ◽  
Brain Functions ◽  
Wide Range ◽  
Gamma Rhythms ◽  
Phase Phase

Oscillations are a hallmark of neural population activity in various brain regions with a spectrum covering a wide range of frequencies. Within this spectrum gamma oscillations have received particular attention due to their ubiquitous nature and to their correlation with higher brain functions. Recently, it has been reported that gamma oscillations in the hippocampus of behaving rodents are segregated in two distinct frequency bands: slow and fast. These two gamma rhythms correspond to different states of the network, but their origin has been not yet clarified. Here, we show theoretically and numerically that a single inhibitory population can give rise to coexisting slow and fast gamma rhythms corresponding to collective oscillations of a balanced spiking network. The slow and fast gamma rhythms are generated via two different mechanisms: the fast one being driven by the coordinated tonic neural firing and the slow one by endogenous fluctuations due to irregular neural activity. We show that almost instantaneous stimulations can switch the collective gamma oscillations from slow to fast and vice versa. Furthermore, to make a closer contact with the experimental observations, we consider the modulation of the gamma rhythms induced by a slower (theta) rhythm driving the network dynamics. In this context, depending on the strength of the forcing and the noise amplitude, we observe phase-amplitude and phase-phase coupling between the fast and slow gamma oscillations and the theta forcing. Phase-phase coupling reveals on average different theta-phases preferences for the two coexisting gamma rhythms joined to a wide cycle-to-cycle variability.

Errant ensembles: dysfunctional neuronal network dynamics in schizophrenia

2010 ◽  
Vol 38 (2) ◽  
pp. 516-521 ◽  
Author(s):  
Matt W. Jones
Keyword(s):  
Calcium Binding ◽  
Field Potential ◽  
Brain Regions ◽  
Aminobutyric Acid ◽  
Neural Population ◽  
Population Activity ◽  
Potential Oscillations ◽  
Disconnection Syndrome ◽  
Hyperactivity Disorder ◽  
Neuronal Network Dynamics

Most complex psychiatric disorders cannot be explained by pathology of a single brain region, but arise as a consequence of dysfunctional interactions between brain regions. Schizophrenia, in particular, has been described as a ‘disconnection syndrome’, but similar principles are likely to apply to depression and ADHD (attention deficit hyperactivity disorder). All these diseases are associated with impaired co-ordination of neural population activity, which manifests as abnormal EEG (electroencephalogram) and LFP (local field potential) oscillations both within and across subcortical and cortical brain regions. Importantly, it is increasingly possible to link oscillations and interactions at distinct frequencies to the physiology and/or pathology of distinct classes of neurons and interneurons. Such analyses increasingly implicate abnormal levels, timing or modulation of GABA (γ-aminobutyric acid)-ergic inhibition in brain disease. The present review discusses the evidence suggesting that dysfunction of a particular class of interneurons, marked by their expression of the calcium-binding protein parvalbumin, could contribute to the broad range of neurophysiological and behavioural symptoms characteristic of schizophrenia.

scholarly journals Coupling of pupil- and neuronal population dynamics reveals diverse influences of arousal on cortical processing

2021 ◽  
Author(s):  
Thomas Pfeffer ◽  
Christian Keitel ◽  
Daniel S. Kluger ◽  
Anne Keitel ◽  
Alena Russmann ◽  
...  
Keyword(s):  
Low Frequency ◽  
Intermediate Frequency ◽  
Brain Regions ◽  
Neuronal Population ◽  
Neural Population ◽  
Cortical Circuits ◽  
Population Activity ◽  
Linear Relationships ◽  
Cortical Regions ◽  
Aperiodic Component

Fluctuations in arousal, controlled by subcortical neuromodulatory systems, continuously shape cortical state, with profound consequences for information processing. Yet, how arousal signals influence cortical population activity in detail has only been characterized for a few selected brain regions so far. Traditional accounts conceptualize arousal as a homogeneous modulator of neural population activity across the cerebral cortex. Recent insights, however, point to a higher specificity of arousal effects on different components of neural activity and across cortical regions. Here, we provide a comprehensive account of the relationships between fluctuations in arousal and neuronal population activity across the human brain. Exploiting the established link between pupil size and central arousal systems, we performed concurrent magnetoencephalographic (MEG) and pupillographic recordings in a large number of participants, pooled across three laboratories. We found a cascade of effects relative to the peak timing of spontaneous pupil dilations: Decreases in low-frequency (2-8 Hz) activity in temporal and lateral frontal cortex, followed by increased high-frequency (>64 Hz) activity in mid-frontal regions, followed by linear and non-linear relationships with intermediate frequency-range activity (8-32 Hz) in occipito-parietal regions. The non-linearity resembled an inverted U-shape whereby intermediate pupil sizes coincided with maximum 8-32 Hz activity. Pupil-linked arousal also coincided with widespread changes in the structure of the aperiodic component of cortical population activity, indicative of changes in the excitation-inhibition balance in underlying microcircuits. Our results provide a novel basis for studying the arousal modulation of cognitive computations in cortical circuits.

Midfrontal conflict-related theta-band power reflects neural oscillations that predict behavior

2013 ◽  
Vol 110 (12) ◽  
pp. 2752-2763 ◽  
Author(s):  
Michael X Cohen ◽  
Tobias H. Donner
Keyword(s):  
Specific Activity ◽  
Physiological Mechanism ◽  
Event Related Potentials ◽  
Brain Regions ◽  
Neural Population ◽  
Neural Oscillations ◽  
Theta Band ◽  
Population Activity ◽  
Theta Power ◽  
Related Potentials

Action monitoring and conflict resolution require the rapid and flexible coordination of activity in multiple brain regions. Oscillatory neural population activity may be a key physiological mechanism underlying such rapid and flexible network coordination. EEG power modulations of theta-band (4–8 Hz) activity over the human midfrontal cortex during response conflict have been proposed to reflect neural oscillations that support conflict detection and resolution processes. However, it has remained unclear whether this frequency-band-specific activity reflects neural oscillations or nonoscillatory responses (i.e., event-related potentials). Here, we show that removing the phase-locked component of the EEG did not reduce the strength of the conflict-related modulation of the residual (i.e., non-phase-locked) theta power over midfrontal cortex. Furthermore, within-subject regression analyses revealed that the non-phase-locked theta power was a significantly better predictor of the conflict condition than was the time-domain phase-locked EEG component. Finally, non-phase-locked theta power showed robust and condition-specific (high- vs. low-conflict) cross-trial correlations with reaction time, whereas the phase-locked component did not. Taken together, our results indicate that most of the conflict-related and behaviorally relevant midfrontal EEG signal reflects a modulation of ongoing theta-band oscillations that occurs during the decision process but is not phase-locked to the stimulus or to the response.

scholarly journals Movement Decomposition in the Primary Motor Cortex

2018 ◽  
Vol 29 (4) ◽  
pp. 1619-1633 ◽  
Author(s):  
Naama Kadmon Harpaz ◽  
David Ungarish ◽  
Nicholas G Hatsopoulos ◽  
Tamar Flash
Keyword(s):  
Motor Cortex ◽  
Primary Motor Cortex ◽  
Temporal Dynamics ◽  
Single Cells ◽  
Brain Regions ◽  
Specific Pattern ◽  
Neural Representation ◽  
Neural Population ◽  
Population Activity ◽  
Acceleration And Deceleration

Abstract A complex action can be described as the composition of a set of elementary movements. While both kinematic and dynamic elements have been proposed to compose complex actions, the structure of movement decomposition and its neural representation remain unknown. Here, we examined movement decomposition by modeling the temporal dynamics of neural populations in the primary motor cortex of macaque monkeys performing forelimb reaching movements. Using a hidden Markov model, we found that global transitions in the neural population activity are associated with a consistent segmentation of the behavioral output into acceleration and deceleration epochs with directional selectivity. Single cells exhibited modulation of firing rates between the kinematic epochs, with abrupt changes in spiking activity timed with the identified transitions. These results reveal distinct encoding of acceleration and deceleration phases at the level of M1, and point to a specific pattern of movement decomposition that arises from the underlying neural activity. A similar approach can be used to probe the structure of movement decomposition in different brain regions, possibly controlling different temporal scales, to reveal the hierarchical structure of movement composition.

scholarly journals Localization of Scattering Objects Using Neural Networks

Sensors ◽  
2020 ◽  
Vol 21 (1) ◽  
pp. 11
Author(s):  
Domonkos Haffner ◽  
Ferenc Izsák
Keyword(s):  
Neural Networks ◽  
Plane Waves ◽  
Measurement Data ◽  
Scattered Wave ◽  
Training Data ◽  
Data Set ◽  
Main Compound ◽  
Incident Plane ◽  
The Neural Networks ◽  
Simulation Package

The localization of multiple scattering objects is performed while using scattered waves. An up-to-date approach: neural networks are used to estimate the corresponding locations. In the scattering phenomenon under investigation, we assume known incident plane waves, fully reflecting balls with known diameters and measurement data of the scattered wave on one fixed segment. The training data are constructed while using the simulation package μ-diff in Matlab. The structure of the neural networks, which are widely used for similar purposes, is further developed. A complex locally connected layer is the main compound of the proposed setup. With this and an appropriate preprocessing of the training data set, the number of parameters can be kept at a relatively low level. As a result, using a relatively large training data set, the unknown locations of the objects can be estimated effectively.

scholarly journals Generation of geometric interpolations of building types with deep variational autoencoders

2020 ◽  
Vol 6 ◽  
Author(s):  
Jaime de Miguel Rodríguez ◽  
Maria Eugenia Villafañe ◽  
Luka Piškorec ◽  
Fernando Sancho Caparrini
Keyword(s):  
Machine Learning ◽  
Neural Networks ◽  
Large Data ◽  
Learning Model ◽  
Large Data Sets ◽  
Data Sets ◽  
Connectivity Map ◽  
Data Set ◽  
3D Objects ◽  
Machine Learning Model

Abstract This work presents a methodology for the generation of novel 3D objects resembling wireframes of building types. These result from the reconstruction of interpolated locations within the learnt distribution of variational autoencoders (VAEs), a deep generative machine learning model based on neural networks. The data set used features a scheme for geometry representation based on a ‘connectivity map’ that is especially suited to express the wireframe objects that compose it. Additionally, the input samples are generated through ‘parametric augmentation’, a strategy proposed in this study that creates coherent variations among data by enabling a set of parameters to alter representative features on a given building type. In the experiments that are described in this paper, more than 150 k input samples belonging to two building types have been processed during the training of a VAE model. The main contribution of this paper has been to explore parametric augmentation for the generation of large data sets of 3D geometries, showcasing its problems and limitations in the context of neural networks and VAEs. Results show that the generation of interpolated hybrid geometries is a challenging task. Despite the difficulty of the endeavour, promising advances are presented.

scholarly journals Determination of Body Parts in Holstein Friesian Cows Comparing Neural Networks and k Nearest Neighbour Classification

Animals ◽  
2020 ◽  
Vol 11 (1) ◽  
pp. 50
Author(s):  
Jennifer Salau ◽  
Jan Henning Haas ◽  
Wolfgang Junge ◽  
Georg Thaller
Keyword(s):  
Machine Learning ◽  
Neural Networks ◽  
Body Parts ◽  
Nearest Neighbour ◽  
Data Set ◽  
3D Data ◽  
Holstein Friesian ◽  
Knn Classification ◽  
Friesian Cows

Machine learning methods have become increasingly important in animal science, and the success of an automated application using machine learning often depends on the right choice of method for the respective problem and data set. The recognition of objects in 3D data is still a widely studied topic and especially challenging when it comes to the partition of objects into predefined segments. In this study, two machine learning approaches were utilized for the recognition of body parts of dairy cows from 3D point clouds, i.e., sets of data points in space. The low cost off-the-shelf depth sensor Microsoft Kinect V1 has been used in various studies related to dairy cows. The 3D data were gathered from a multi-Kinect recording unit which was designed to record Holstein Friesian cows from both sides in free walking from three different camera positions. For the determination of the body parts head, rump, back, legs and udder, five properties of the pixels in the depth maps (row index, column index, depth value, variance, mean curvature) were used as features in the training data set. For each camera positions, a k nearest neighbour classifier and a neural network were trained and compared afterwards. Both methods showed small Hamming losses (between 0.007 and 0.027 for k nearest neighbour (kNN) classification and between 0.045 and 0.079 for neural networks) and could be considered successful regarding the classification of pixel to body parts. However, the kNN classifier was superior, reaching overall accuracies 0.888 to 0.976 varying with the camera position. Precision and recall values associated with individual body parts ranged from 0.84 to 1 and from 0.83 to 1, respectively. Once trained, kNN classification is at runtime prone to higher costs in terms of computational time and memory compared to the neural networks. The cost vs. accuracy ratio for each methodology needs to be taken into account in the decision of which method should be implemented in the application.

Sign in / Sign up

Export Citation Format

Share Document