Feedforward Processing Research Articles

Differential contributions of the C1 ERP and Broadband high-frequency activity to Visual Processing.

The high-frequency activity (HFA, 80 - 150 Hz) in human intracranial recordings shows a differential modulation to different degrees in contrast when stimuli are behaviorally relevant, indicating a feedforward process. However, the HFA is also significantly dominated by superficial layers and exhibits a peak before 200 ms, suggesting that it is more likely a feedback signal. MEG recordings are suited to reveal an HFA modulation similar to its modulation in intracranial recordings. This allows for non-invasive, direct comparison of HFA with the C1, an established measure for feedforward input to V1, to test whether HFA represents feedforward or rather feedback. In simultaneous recordings, we used the EEG-C1 ERP component and MEG-HFA to define feedforward processing in visual cortices. C1 latency preceded the HFA peak modulation, which had a more sustained response. Furthermore, modulation parameters like onset, peak time and peak amplitude were uncorrelated. Most importantly, the C1 but not HFA distinguished small task irrelevant contrast differences in visual stimulation. These results highlight the differential roles for the C1 and HFA in visual processing with the C1 measuring feedforward discrimination ability and HFA indexing feedforward and feedback processing.

Sensitivity of Spiking Neural Networks Due to Input Perturbation.

Background: To investigate the behavior of spiking neural networks (SNNs), the sensitivity of input perturbation serves as an effective metric for assessing the influence on the network output. However, existing methods fall short in evaluating the sensitivity of SNNs featuring biologically plausible leaky integrate-and-fire (LIF) neurons due to the intricate neuronal dynamics during the feedforward process. Methods: This paper first defines the sensitivity of a temporal-coded spiking neuron (SN) as the deviation between the perturbed and unperturbed output under a given input perturbation with respect to overall inputs. Then, the sensitivity algorithm of an entire SNN is derived iteratively from the sensitivity of each individual neuron. Instead of using the actual firing time, the desired firing time is employed to derive a more precise analytical expression of the sensitivity. Moreover, the expectation of the membrane potential difference is utilized to quantify the magnitude of the input deviation. Results/Conclusions: The theoretical results achieved with the proposed algorithm are in reasonable agreement with the simulation results obtained with extensive input data. The sensitivity also varies monotonically with changes in other parameters, except for the number of time steps, providing valuable insights for choosing appropriate values to construct the network. Nevertheless, the sensitivity exhibits a piecewise decreasing tendency with respect to the number of time steps, with the length and starting point of each piece contingent upon the specific parameter values of the neuron.

A no-reference video quality assessment method with bidirectional hierarchical semantic representation

Perceived quality assessment for user-generated content (UGC) videos is of considerable importance to safeguard the viewing experience of end-users. The diversity of content and the blend of authentic distortions pose great challenges for UGC video quality assessment (UGC-VQA). The reverse hierarchy theory suggests that there is bottom-up feedforward perception and top-down feedback perception in the human visual system (HVS). However, existing UGC-VQA methods rarely consider feedback perception and make it difficult to model the complete visual perception loop, leading to inaccurate prediction of perceived quality. Thus, this paper innovatively proposes a bidirectional hierarchical semantic extraction structure for VQA (BHSE-VQA), which simulates visual feedforward and feedback perception. Specifically, a feedforward and feedback multi-level network is first designed to extract multi-level spatio-temporal features with a 3D-ConvNext backbone in the feedforward pathway and process these hierarchical features with the combination of channel attention and spatial attention mechanisms in the feedback pathway. Then, considering the varying impacts of responses at different perception layers on visual perception results, the weights of features at each level are redistributed to be consistent with human perception. With bidirectional hierarchical pathway features, a temporal attention fusion network is introduced to further capture temporal correlations and aggregate features relying on the residuals between feedforward and feedback features for quality prediction. Experimental results on several representative UGC-VQA databases demonstrate the effectiveness of the proposed model and the significance of comprehensive hierarchical perception modeling for UGC-VQA.

Differential neural mechanisms underlie cortical gating of visual spatial attention mediated by alpha-band oscillations

Selective attention relies on neural mechanisms that facilitate processing of behaviorally relevant sensory information while suppressing irrelevant information, consistently linked to alpha-band oscillations in human M/EEG studies. We analyzed cortical alpha responses from intracranial electrodes implanted in eight epilepsy patients, who performed a visual spatial attention task. Electrocorticographic data revealed a spatiotemporal dissociation between attention-modulated alpha desynchronization, associated with the enhancement of sensory processing, and alpha synchronization, associated with the suppression of sensory processing, during the cue-target interval. Dorsal intraparietal areas contralateral to the attended hemifield primarily exhibited a delayed and sustained alpha desynchronization, while ventrolateral extrastriatal areas ipsilateral to the attended hemifield primarily exhibited an earlier and sustained alpha synchronization. Analyses of cross-frequency coupling between alpha phase and broadband high-frequency activity (HFA) further revealed cross-frequency interactions along the visual hierarchy contralateral to the attended locations. Directionality analyses indicate that alpha phase in early and extrastriatal visual areas modulated HFA power in downstream visual areas, thus potentially facilitating the feedforward processing of an upcoming, spatially predictable target. In contrast, in areas ipsilateral to the attended locations, HFA power modulated local alpha phase in early and extrastriatal visual areas, with suppressed interareal interactions, potentially attenuating the processing of distractors. Our findings reveal divergent alpha-mediated neural mechanisms underlying target enhancement and distractor suppression during the deployment of spatial attention, reflecting enhanced functional connectivity at attended locations, while suppressed functional connectivity at unattended locations. The collective dynamics of these alpha-mediated neural mechanisms play complementary roles in the efficient gating of sensory information.

Reality and imagination intertwined: A sensorimotor paradox interpretation

As a hypothesis on the origins of mind and language, the evolutionary theory of the sensorimotor paradox suggests that capacities for imagination, self-representation and abstraction would operate from a dissociation in what is known as the forward model. In some studies, sensory perception is understood as a system of prediction and confirmation (feedforward and feedback processes) that would share common yet distinct and overlapping neural networks with mental imagery. The latter would then mostly operate through internal feedback processes. The hypothesis of our theory is that dissociation and parallelism between those processes would make it less likely for imaginary prediction to match and simultaneously coincide with any sensory feedback, contradicting the stimulus/response pattern. The gap between the two and the effort required to maintain this gap, born from the development of bipedal stance and a radical change to our relation to our own hands, would be the very structural foundation to our capacity to elaborate abstract thoughts, by partially blocking and inhibiting motor action. Mental imagery would structurally be dissociated from perception, though maintaining an intricated relation of interdependence. Moreover, the content of the images would be subordinate to their function as emotional regulators, prioritising consistency with some global, conditional and socially learnt body-image. As a higher-level and proto-aesthetic function, we can speculate that the action and instrumentalisation of dissociating imagination from perception would become the actual prediction and their coordination, the expected feedback.

Development of a doubly-fed compound control for hybrid ground source heat pump systems in hot summer and cold winter areas

In hot summer and cold winter areas, the impact of soil thermal accumulation is significant even for hybrid ground source heat pump systems (HGSHPs), especially after a long-term operation. To alleviate the energy performance degradation caused by soil temperature rise, a detailed temperature field heat transfer model of ground heat exchangers (GHXs) has been established and a doubly-fed compound control for HGSHPs is developed and integrated into a central controller module. The central controller can automatically control the loop flow distribution of units and achieve group activation of GHXs based on the load sequences forecasting feedforward and system variables feedback processes. Five control modes are included in the central controller, and the energy performances and soil temperature rise distribution of the HGSHPs with different control modes for fifteen years are compared. The results indicated that the HGSHPs performs best in terms of energy performance and soil thermal alleviation with Control Unit 13. Compared to the Normal control mode, the mean water temperature of GHXs, energy consumption of the chiller and heat pump decreased by 1.6 °C, 5.9 %, and 15.5 %, respectively, and the units COP increased by 11.3 %.

Feedback needs experience

Visual perception requires aligned feedforward and feedback processing, yet the role of experience remains unclear. The study by Dias etal.1 in this issue of Neuron shows that the retinotopic organization of orientation-tuned feedback from higher to primary visual cortex is learned in mice.

Dynamic decoupling and compensation based on the inverted network technique and experiment research for multi-axis hydraulic road simulator

This paper presents a dynamic decoupling and compensation approach to eliminate channel interaction for a multi-axis hydraulic road simulator. Misalignment of the concentrated moving mass centroid on the entire test bench with the control point and inconsistencies in actuator dynamics leads to dynamic coupling, which deteriorates the motion performance of the system. To solve the problem, an inverted decoupling and compensation network based on an identification model is proposed. Firstly, the system is modeled using the recursive extended least square (RELS) identification method. In the feedforward compensation modeling process, the steady-state inverse is obtained using zero-magnitude error technology control (ZMETC). In addition, a finite impulse response (FIR) filter with an adaptive algorithm is employed to reduce the adverse impact of the modeling error. The proposed decoupling strategy is validated by conducting experiments using a multi-axis hydraulic road simulator. The experimental results indicate that the designed compensator can effectively decrease the cross-coupling of multi-input and multi-output (MIMO) systems. The suggested approach is also applicable to applications that necessitate the elimination of interactions between different control variables.

NMDA receptor antagonist high-frequency oscillations are transmitted via bottom-up feedforward processing

In mammals, NMDA receptor antagonists have been linked to the emergence of high-frequency oscillations (HFO, 130–180 Hz) in cortical and subcortical brain regions. The extent to which transmission of this rhythm is dependent on feedforward (bottom-up) or feedback (top-down) mechanisms is unclear. Previously, we have shown that the olfactory bulb (OB), known to orchestrate oscillations in many brain regions, is an important node in the NMDA receptor-dependent HFO network. Since the piriform cortex (PC) receives major input from the OB, and can modulate OB activity via feedback projections, it represents an ideal site to investigate transmission modalities. Here we show, using silicon probes, that NMDA receptor antagonist HFO are present in the PC associated with current dipoles, although of lower power than the OB. Granger causality and peak-lag analyses implicated the OB as the driver of HFO in the PC. Consistent with this, reversible inhibition of the OB resulted in a reduction of HFO power both locally and in the PC. In contrast, inhibition of the PC had minimal impact on OB activity. Collectively, these findings point to bottom-up mechanisms in mediating the transmission of NMDA receptor antagonist-HFO, at least in olfactory circuits.

Role of proprioception in corrective visually-guided movements: larger movement errors in both arms of a deafferented individual compared to control participants.

Proprioception plays an important role in both feedforward and feedback processes underlying movement control. This has been shown with individuals who suffered a profound proprioceptive loss and use vision to partially compensate for the sensory loss. The purpose of this study was to specifically examine the role of proprioception in feedback motor responses to visual perturbations by examining voluntary arm movements in an individual with a rare case of selective peripheral deafferentation (GL). We compared her left and right hand movements with those of age-matched female control participants (70.0 years ± 0.2 SEM) during a reaching task. Participants were asked to move their unseen hand, represented by a cursor on the screen, quickly and accurately to reach a visual target. A visual perturbation could be pseudorandomly applied, at movement onset, to either the target position (target jump) or the cursor position (cursor jump). Results showed that despite the continuous visual feedback that was provided, GL produced larger errors in final position accuracy compared to control participants, with her left nondominant hand being more erroneous after a cursor jump. We also found that the proprioceptively-deafferented individual produced less spatially efficient movements than the control group. Overall, these results provide evidence of a heavier reliance on proprioceptive feedback for movements of the nondominant hand relative to the dominant hand, supporting the view of a lateralization of the feedback processes underlying motor control.

Mouse adaptation of human inflammatory bowel diseases microbiota enhances colonization efficiency and alters microbiome aggressiveness depending on the recipient colonic inflammatory environment

BackgroundUnderstanding the cause vs consequence relationship of gut inflammation and microbial dysbiosis in inflammatory bowel diseases (IBD) requires a reproducible mouse model of human-microbiota-driven experimental colitis.ResultsOur study demonstrated that human fecal microbiota transplant (FMT) transfer efficiency is an underappreciated source of experimental variability in human microbiota-associated (HMA) mice. Pooled human IBD patient fecal microbiota engrafted germ-free (GF) mice with low amplicon sequence variant (ASV)-level transfer efficiency, resulting in high recipient-to-recipient variation of microbiota composition and colitis severity in HMA Il-10−/− mice. In contrast, mouse-to-mouse transfer of mouse-adapted human IBD patient microbiota transferred with high efficiency and low compositional variability resulting in highly consistent and reproducible colitis phenotypes in recipient Il-10−/− mice. Engraftment of human-to-mouse FMT stochastically varied with individual transplantation events more than mouse-adapted FMT. Human-to-mouse FMT caused a population bottleneck with reassembly of microbiota composition that was host inflammatory environment specific. Mouse-adaptation in the inflamed Il-10−/− host reassembled a more aggressive microbiota that induced more severe colitis in serial transplant to Il-10−/− mice than the distinct microbiota reassembled in non-inflamed WT hosts.ConclusionsOur findings support a model of IBD pathogenesis in which host inflammation promotes aggressive resident bacteria, which further drives a feed-forward process of dysbiosis exacerbated by gut inflammation. This model implies that effective management of IBD requires treating both the dysregulated host immune response and aggressive inflammation-driven microbiota. We propose that our mouse-adapted human microbiota model is an optimized, reproducible, and rigorous system to study human microbiome-driven disease phenotypes, which may be generalized to mouse models of other human microbiota-modulated diseases, including metabolic syndrome/obesity, diabetes, autoimmune diseases, and cancer.9-8iLEe2BorBmu_z2q_8UWVideo

Neural tracking of speech acoustics in noise is coupled with lexical predictability as estimated by large language models.

Adults heard recordings of two spatially separated speakers reading newspaper and magazine articles. They were asked to listen to one of them and ignore the other, and EEG was recorded to assess their neural processing. Machine learning extracted neural sources that tracked the target and distractor speakers at three levels: the acoustic envelope of speech (delta- and theta-band modulations), lexical frequency for individual words, and the contextual predictability of individual words estimated by GPT-4 and earlier lexical models. To provide a broader view of speech perception, half of the subjects completed a simultaneous visual task, and the listeners included both native and non-native English speakers. Distinct neural components were extracted for these levels of auditory and lexical processing, demonstrating that native English speakers had greater target-distractor separation compared to non-native English speakers on most measures, and that lexical processing was reduced by the visual task. Moreover, there was a novel interaction of lexical predictability and frequency with auditory processing; acoustic tracking was stronger for lexically harder words, suggesting that people listened harder to the acoustics when needed for lexical selection. This demonstrates that speech perception is not simply a feedforward process from acoustic processing to the lexicon. Rather, the adaptable context-sensitive processing long known to occur at a lexical level has broader consequences for perception, coupling with the acoustic tracking of individual speakers in noise.Significance Statement In challenging listening conditions, people use focused attention to help understand individual talkers and ignore others, which changes their neural processing for speech at auditory through lexical levels. However, lexical processing for natural materials (e.g., conversations, audiobooks) has been difficult to measure, because of limitations of tools to estimate the predictability of individual words in longer discourses. The present investigation uses a contemporary large language model, GPT-4, to estimate word predictability, and demonstrates that listeners make online adaptations to their auditory neural processing in accord with these predictions; neural activity more closely tracks the acoustics of the target talker when words are less predictable from the context.

Towards a unified ontology of cybernetics: bridging mechanology and CAS for cooperative societies through thematic synthesis

PurposeThis paper aims to conceptually unite an ontology of cybernetics, bridging living and technical systems, to facilitate future epistemological and theoretical advancements applicable to highly technical societies by crafting a set of definitions that elucidate the nature of the world in which these systems operate.Design/methodology/approachThe research uses a thematic synthesis of two systems/cybernetic traditions: complex adaptive systems and mechanology. The primary sources for this research were the main theses and correlated papers published in the Simondon case for mechanology, and the seminal literature preselected by the Santa Fe Institute for complex adaptive systems.FindingsThe study proposes the following concepts: Individuation: the emergence of new properties in an individual composed of synergistically related parts; Technical evolution: the notion that technical objects evolve into living beings; circular causality: the notion that feedback and feedforward processes shape the organisation and structure of systems and their relationship with the environment; The milieu refers to the part of the environment that has a relationship of co-production, co-dependency, and co-evolution with systems. Metastability is a state that transcends stability and instability and motivates changes in the system. Transduction is the cumulative process of individuation in which systems change structure and organization to maintain operational coherence with their surroundings.Originality/valueThe concepts the paper identifies can serve as a starting point for an extended study on the ontology of cybernetics or as the basis for an evolutionary epistemology both in humans and machines.

STDCformer: Spatial-temporal dual-path cross-attention model for fMRI-based autism spectrum disorder identification

Resting-state functional magnetic resonance imaging (rs-fMRI) is a non-invasive neuroimaging technique widely utilized in the research of Autism Spectrum Disorder (ASD), providing preliminary insights into the potential biological mechanisms underlying ASD. Deep learning techniques have demonstrated significant potential in the analysis of rs-fMRI. However, accurately distinguishing between healthy control group and ASD has been a longstanding challenge. In this regard, this work proposes a model featuring a dual-path cross-attention framework for spatial and temporal patterns, named STDCformer, aiming to enhance the accuracy of ASD identification. STDCformer can preserve both temporal-specific patterns and spatial-specific patterns while explicitly interacting spatiotemporal information in depth. The embedding layer of the STDCformer embeds temporal and spatial patterns in dual paths. For the temporal path, we introduce a perturbation positional encoding to improve the issue of signal misalignment caused by individual differences. For the spatial path, we propose a correlation metric based on Gramian angular field similarity to establish a more specific whole-brain functional network. Subsequently, we interleave the query and key vectors of dual paths to interact spatial and temporal information. We further propose integrating the dual-path attention into a tensor that retains spatiotemporal dimensions and utilizing 2D convolution for feed-forward processing. Our attention layer allows the model to represent spatiotemporal correlations of signals at multiple scales to alleviate issues of information distortion and loss. Our STDCformer demonstrates competitive results compared to state-of-the-art methods on the ABIDE dataset. Additionally, we conducted interpretative analyses of the model to preliminarily discuss the potential physiological mechanisms of ASD. This work once again demonstrates the potential of deep learning technology in identifying ASD and developing neuroimaging biomarkers for ASD.

Artificial Neural Network Model Application for Validating and Predicting Ruminant Population in East Java, Indonesia

Abstract The aim of this study was to develop a model that could estimate the population of ruminants (beef cattle, goats, and sheep) in East Java with high accuracy. Secondary data on ruminant livestock populations in 31 regencies in East Java Province were acquired from the Central Bureau of Statistics of East Java between 2009 and 2020. Modeling backpropagation neural networks and feedforward learning processes with architectural models for beef cow population 11-3-1, goat population 11-3-1, and sheep population 11-5-1 are used to predict the population of ruminants in East Java. This architecture has a precision of more than 90%. The prediction RSquared (R2) for the beef cattle population is 0.997, the goat population is 0.998, and the sheep population is 0.975.

The attentive reconstruction of objects facilitates robust object recognition.

Humans are extremely robust in our ability to perceive and recognize objects-we see faces in tea stains and can recognize friends on dark streets. Yet, neurocomputational models of primate object recognition have focused on the initial feed-forward pass of processing through the ventral stream and less on the top-down feedback that likely underlies robust object perception and recognition. Aligned with the generative approach, we propose that the visual system actively facilitates recognition by reconstructing the object hypothesized to be in the image. Top-down attention then uses this reconstruction as a template to bias feedforward processing to align with the most plausible object hypothesis. Building on auto-encoder neural networks, our model makes detailed hypotheses about the appearance and location of the candidate objects in the image by reconstructing a complete object representation from potentially incomplete visual input due to noise and occlusion. The model then leverages the best object reconstruction, measured by reconstruction error, to direct the bottom-up process of selectively routing low-level features, a top-down biasing that captures a core function of attention. We evaluated our model using the MNIST-C (handwritten digits under corruptions) and ImageNet-C (real-world objects under corruptions) datasets. Not only did our model achieve superior performance on these challenging tasks designed to approximate real-world noise and occlusion viewing conditions, but also better accounted for human behavioral reaction times and error patterns than a standard feedforward Convolutional Neural Network. Our model suggests that a complete understanding of object perception and recognition requires integrating top-down and attention feedback, which we propose is an object reconstruction.

Loss of postnatal Arx transcriptional activity in parvalbumin interneurons reveals non-cell autonomous disturbances in CA1 pyramidal cells

Maintenance of proper electrophysiological and connectivity profiles in the adult brain may be a perturbation point in neurodevelopmental disorders (NDDs). How these profiles are maintained within mature circuits is unclear. We recently demonstrated that postnatal ablation of the Aristaless (Arx) homeobox gene in parvalbumin interneurons (PVIs) alone led to dysregulation of their transcriptome and alterations in their functional as well as network properties in the hippocampal cornu Ammoni first region (CA1). Here, we characterized CA1 pyramidal cells (PCs) responses in this conditional knockout (CKO) mouse to further understand the circuit mechanisms by which postnatal Arx expression regulates mature CA1 circuits. Field recordings of network excitability showed that CA1 PC ensembles were less excitable in response to unpaired stimulations but exhibited enhanced excitability in response to paired-pulse stimulations. Whole-cell voltage clamp recordings revealed a significant increase in the frequency of spontaneous inhibitory postsynaptic currents onto PCs. In contrast, excitatory drive from evoked synaptic transmission was reduced while that of inhibitory synaptic transmission was increased. Current clamp recordings showed increase excitability in several sub- and threshold membrane properties that correlated with an increase in voltage-gated Na+ current. Our data suggest that, in addition to cell-autonomous disruption in PVIs, loss of Arx postnatal transcriptional activity in PVIs led to complex dysfunctions in PCs in CA1 microcircuits. These non-cell autonomous effects are likely the product of breakdown in feedback and/or feedforward processes and should be considered as fundamental contributors to the circuit mechanisms of NDDs such as Arx-linked early-onset epileptic encephalopathies.

Neural oscillations guiding action during effects imagery

Goal-directed acting requires the integration of sensory information but can also be performed without direct sensory input. Examples of this can be found in sports and can be conceptualized by feedforward processes. There is, however, still a lack of understanding of the temporal neural dynamics and neuroanatomical structures involved in such processes. In the current study, we used EEG beamforming methods and examined 37 healthy participants in two well-controlled experiments varying the necessity of anticipatory processes during goal-directed action. We found that alpha and beta activity in the medial and posterior cingulate cortex enabled feedforward predictions about the position of an object based on the latest sensorimotor state. On this basis, theta band activity seems more related to sensorimotor representations, while beta band activity would be more involved in setting up the structure of the neural representations themselves. Alpha band activity in sensory cortices reflects an intensified gating of the anticipated perceptual consequences of the to-be-executed action. Together, the findings indicate that goal-directed acting through the anticipation of the predicted state of an effector is based on accompanying processes in multiple frequency bands in midcingulate and sensory brain regions.

Do They Know It's Christmash? Lexical Knowledge Directly Impacts Speech Perception.

We recently reported strong, replicable (i.e., replicated) evidence for lexically mediated compensation for coarticulation (LCfC; Luthra etal., 2021), whereby lexical knowledge influences a prelexical process. Critically, evidence for LCfC provides robust support for interactive models of cognition that include top-down feedback and is inconsistent with autonomous models that allow only feedforward processing. McQueen, Jesse, and Mitterer (2023) offer five counter-arguments against our interpretation; we respond to each of those arguments here and conclude that top-down feedback provides the most parsimonious explanation of extant data.

Effects of altered sensory feedback on piano performance errors: An exploratory study

Music performance is an intensive sensorimotor task that involves the generation of mental representations of musical information that are actively accessed, maintained, and manipulated according to the demands of the performance. Internal representations and external information interact through feedback and feedforward processes that adjust the musician’s motor behavior to optimize a musical performance. This study aimed to examine the relationship between altered sensory feedback and performance errors. Seventeen experienced pianists aged between 33 and 54 years performed Hanon Exercise N°1 from memory under four different conditions: (1) normal (normal sensory feedback); (2) closed fallboard (altered haptic and auditory feedback); (3) blindfolded (altered visual feedback); and (4) combined (blindfolded and closed fallboard; altered haptic, auditory, and visual feedback). Performance errors were quantified based on a video analysis of the performances. Results indicated that compared with normal performance, participants made significantly more note errors in the blindfolded condition and more bar-adding errors per trial in the closed fallboard condition. The comparison between the normal condition and the three altered sensory feedback conditions revealed the impact of altering sensory feedback in musical performance. These findings are discussed in the context of music learning.