Stable Representation Research Articles

An Integrated Design Scheme for SKR-Based Data-Driven Dynamic Fault Detection Systems

In this article, an integrated design diagram for a stable kernel representation (SKR)-based data-driven fault detection (FD) system and performance criteria is proposed for stochastic dynamic systems in the probabilistic sense. A new distributionally robust FD system is developed using input and output data in the absence of a system model and perfect probability distributions for noises and random faults. To be specific, an SKR-based data-driven primary residual generator is first constructed. By introducing the so-called mean-covariance based ambiguity sets, families of probability distributions of the primary residual in fault-free and the concerned multiple faulty cases are characterized. The FD system design is then formulated as a distributionally robust optimization problem in the sense of minimizing the missed detection rate (MDR) with a predefined upper bound of false alarm rate (FAR). With the aid of worst-case conditional value-at-risk, a matrix-valued distribution independent solution to the targeting FD problem is derived without posing specific distribution assumptions. The developed FD system is, thus, robust against the distributional uncertainties of noises and random faults. Simultaneously, a tighter upper bound of MDR for an identical FAR criterion is achieved in comparison with the vector-valued distributionally robust FD method. An experimental study on a laboratory setup of a three-tank system shows the applicability of the proposed method.

Is there evidence for a noisy computation deficit in developmental dyslexia?

The noisy computation hypothesis of developmental dyslexia (DD) is particularly appealing because it can explain deficits across a variety of domains, such as temporal, auditory, phonological, visual and attentional processes. A key prediction is that noisy computations lead to more variable and less stable word representations. A way to test this hypothesis is through repetition of words, that is, when there is noise in the system, the neural signature of repeated stimuli should be more variable. The hypothesis was tested in an functional magnetic resonance imaging experiment with dyslexic and typical readers by repeating words twelve times. Variability measures were computed both at the behavioral and neural levels. At the behavioral level, we compared the standard deviation of reaction time distributions of repeated words. At the neural level, in addition to standard univariate analyses and measures of intra-item variability, we also used multivariate pattern analyses (representational similarity and classification) to find out whether there was evidence for noisier representations in dyslexic readers compared to typical readers. Results showed that there were no significant differences between the two groups in any of the analyses despite robust results within each group (i.e., high representational similarity between repeated words, good classification of words vs. non-words). In summary, there was no evidence in favor of the idea that dyslexic readers would have noisier neural representations than typical readers.

Stability of motor representations after paralysis.

Neural plasticity allows us to learn skills and incorporate new experiences. What happens when our lived experiences fundamentally change, such as after a severe injury? To address this question, we analyzed intracortical population activity in the posterior parietal cortex (PPC) of a tetraplegic adult as she controlled a virtual hand through a brain-computer interface (BCI). By attempting to move her fingers, she could accurately drive the corresponding virtual fingers. Neural activity during finger movements exhibited robust representational structure similar to fMRI recordings of able-bodied individuals' motor cortex, which is known to reflect able-bodied usage patterns. The finger representational structure was consistent throughout multiple sessions, even though the structure contributed to BCI decoding errors. Within individual BCI movements, the representational structure was dynamic, first resembling muscle activation patterns and then resembling the anticipated sensory consequences. Our results reveal that motor representations in PPC reflect able-bodied motor usage patterns even after paralysis, and BCIs can re-engage these stable representations to restore lost motor functions.

Flexible cue anchoring strategies enable stable head direction coding in both sighted and blind animals

Vision plays a crucial role in instructing the brain’s spatial navigation systems. However, little is known about how vision loss affects the neuronal encoding of spatial information. Here, recording from head direction (HD) cells in the anterior dorsal nucleus of the thalamus in mice, we find stable and robust HD tuning in rd1 mice, a model of photoreceptor degeneration, that go blind by approximately one month of age. In contrast, placing sighted animals in darkness significantly impairs HD cell tuning. We find that blind mice use olfactory cues to maintain stable HD tuning and that prior visual experience leads to refined HD cell tuning in blind rd1 adult mice compared to congenitally blind animals. Finally, in the absence of both visual and olfactory cues, the HD attractor network remains intact but the preferred firing direction of HD cells drifts over time. These findings demonstrate flexibility in how the brain uses diverse sensory information to generate a stable directional representation of space.

Seasonal Dynamics of Zygnema (Zygnematophyceae) Mats from the Austrian Alps.

Filamentous green algae of the genus Zygnema are an essential part of hydro-terrestrial ecosystems. Despite several studies on their resistance to natural stresses, little is known about the composition of their assemblages and the changes they undergo over time. Two sites at altitudes above 2200m a.s.l. in the Austrian Alps were selected for a 2-year observation period and sampled five times. Molecular phylogenetic analysis of the 152 isolated strains of Zygnema sp. was performed based on the rbcL and trnG sequences. Seven genotypes were found at these sites during the samplings, but their proportion varied throughout the seasons. The site with a more stable water regime also had a more stable representation of genotypes, in contrast to the site with fluctuating water availability. The mats formed resistant pre-akinetes at the end of the season with reduced photosynthetic activity. Contrary to expectations, the mats were not exposed to extremely cold temperatures in winter due to snow cover. Some genotypes have been previously observed at this site, indicating that the population composition is stable. This work highlights the importance of resistant pre-akinetes in surviving winter conditions, the ability of algae to re-establish mats, and the need to address the hidden diversity of the genus Zygnema.

Detection of Malicious Attacks in Autonomous Cyber-Physical Inverter-Based Microgrids

The distributed generation capabilities of microgrids (MGs) arise as essential assets in enhancing grid resilience. The integration of distributed energy sources, controllable loads, and prosumers necessitates the deployment of potent control and communication synergies. While those synergies transform MGs into cyber-physical systems through information technologies able to sense, control, and actuate local resources and loads, they inadvertently expose MGs to cyber-attack threats. Increasing the security of critical communication and control systems against “black swan” events, i.e., high-impact low-probability cyber-physical incidents, is a major priority for MG operations. Such incidents, if left unabated, can intensify and elicit system dynamics instability, eventually causing outages and system failures. In this article, we develop an integrated approach for multiagent MG systems able to perform the detection of malicious cyber-physical attacks based on subspace methods. We employ the small-signal model of an autonomous/islanded MG and consider different attack models targeting the MG’s secondary frequency control. The attack detector is constructed via identifying the stable kernel representation of the autonomous cyber-physical MG in the attack-free case. We illustrate the impact of the attack models as well as the feasibility of the developed detection method in simulation models of the Canadian urban benchmark distribution system.

Computational Topology and its Applications in Geometric Design

Background: In recent geometric design, many effective toolkits for geometric modeling and optimization have been proposed and applied in practical cases, while effective and efficient designing of shapes that have desirable topological properties remains to be a challenge. The development of computational topology, especially persistent homology, permits convenient usage of topological invariants in shape analysis, geometric modeling, and shape optimization. Persistence diagram, the useful topological summary of persistent homology, provides a stable representation of multiscale homology invariants in the presence of noise in original data. Recent works show the wide use of persistent homology tools in geometric design. Objective: In this paper, we review the geometric design based on computational topological tools in three aspects: the extraction of topological features and representations, topology-aware shape modeling, and topology-based shape optimization. Methods: By tracking the development of each aspect and comparing the methods using classical topological invariants, motivations, and key approaches of important related works based on persistent homology are clarified. Results : We review geometric design through topological extraction, topological design, and shape optimization based on topology preservation. Related works show the successful applications of computational topology tools of geometric design. Conclusion: Solutions for the proposed core problems will affect the geometric design and its applications. In the future, the development of computational topology may boost computer-aided topological design.

Contribution of behavioural variability to representational drift.

Neuronal responses to similar stimuli change dynamically over time, raising the question of how internal representations can provide a stable substrate for neural coding. Recent work has suggested a large degree of drift in neural representations even in sensory cortices, which are believed to store stable representations of the external world. While the drift of these representations is mostly characterized in relation to external stimuli, the behavioural state of the animal (for instance, the level of arousal) is also known to strongly modulate the neural activity. We therefore asked how the variability of such modulatory mechanisms can contribute to representational changes. We analysed large-scale recording of neural activity from the Allen Brain Observatory, which was used before to document representational drift in the mouse visual cortex. We found that, within these datasets, behavioural variability significantly contributes to representational changes. This effect was broadcasted across various cortical areas in the mouse, including the primary visual cortex, higher order visual areas, and even regions not primarily linked to vision like hippocampus. Our computational modelling suggests that these results are consistent with independent modulation of neural activity by behaviour over slower timescales. Importantly, our analysis suggests that reliable but variable modulation of neural representations by behaviour can be misinterpreted as representational drift if neuronal representations are only characterized in the stimulus space and marginalized over behavioural parameters.

Stable image representation based stability performance monitoring and recovery for feedback control systems

This paper proposes a scheme to achieve real-time stability performance monitoring (SPM) and designs performance recovery controllers (PRCs) for feedback control systems with multiplicative faults. To be specific, a stability performance indicator is presented with the aid of stable image representation (SIR) for multiplicative faults through coprime factorization techniques. On this basis, a systematically hierarchical SPM and PRC scheme is proposed with three thresholds derived to evaluate the performance degradation degree. Subsequently, an integrated model-based and data-driven SIR-based PRC is presented to recover system stability performance. The embedded PRC parameters are adaptive to system variations by means of identifying the SIR of the faulty plant through a dual parity vector. Besides, some QR-decomposition based data-driven techniques are provided for the implementation of PRC to improve computation efficiency. Finally, the effectiveness of the proposed approach is demonstrated on a boost circuit model.

Weakly positive and directed Anosov representations

Given a finitely generated group Γ, a directed graph Λ, and a map R:Λ→Γ, we introduce the notion of an (R,Λ)-directed Anosov representation. This is a weakening of the notion of Anosov representations. Our main theorem gives a procedure to construct (R,Λ)-directed Anosov representations using Fock-Goncharov positivity. As an application of our main theorem, we construct large families of primitive stable representations from F2 to PGL(V), including non-discrete and non-faithful examples.

Fault Detection Approach for Nonlinear Systems via Nonlinear Factorization and Fuzzy Models

This brief aims to propose an effective fault detection method for general class of nonlinear systems in the context of the closed-loop system stability. To this end, the nonlinear factorization technique is first used to model the faulty nonlinear systems, which can be represented by the so-called stable kernel representation with a stable parameterization of the system changes triggered by faults. Then, the closed-loop system stability is discussed according to the internal stability definition and the small gain theorem, respectively, to present the design framework of the fault detection system. Different from the traditional fault detection schemes, the proposed fault detection approach focuses on detecting whether the system closed-loop stability is damaged by faults utilizing the online measurable system and controller residual signals. Furthermore, for the implementation of the proposed fault detection framework, Takagi-Sugeno fuzzy models are applied to approximate the nonlinear systems and thus the fault detection system design methods can be provided by taking advantage of the linear matrix inequality technique. Finally, a case study is used to verify the achieved results.

Infants’ sensitivity to emotional expressions in actions: The contributions of parental expressivity and motor experience

Actions can convey information about the affective state of an actor. By the end of the first year, infants show sensitivity to such emotional information in actions. Here, we examined the mechanisms contributing to infants’ developing sensitivity to emotional action kinematics. We hypothesized that this sensitivity might rely on two factors: a stable motor representation of the observed action to be able to detect deviations from how it would typically be performed and experience with emotional expressions. The sensitivity of 12- to 13-month-old infants to happy and angry emotional cues in a manual transport action was examined using facial EMG. Infants’ own movements when performing an object transport task were assessed using optical motion capture. The infants’ caregivers’ emotional expressivity was measured using a questionnaire. Negative emotional expressivity of the primary caregiver was significantly related to infants’ sensitivity to observed angry actions. There was no evidence for such an association with infants’ own motor skill. Overall, our results show that infants’ experience with emotions, measured as caregivers’ emotional expressivity, may aid infants’ discrimination of others’ emotions expressed in action kinematics.

Unlabeled data driven cost-sensitive inverse projection sparse representation-based classification with 1/2 regularization

Sparse representation-based classification (SRC) has been widely used because it just relies on simple linear regression ideas to do classification, and it does not need learning. However, the performance of SRC is limited by needing sufficient labeled samples per class and the sensitivity to class imbalance. For tackling these problems, an improved SRC model is constructed in this paper. For alleviating the problem of insufficient labeled samples, an unlabeled data-driven inverse projection sparse representation-based classification model is constructed to achieve effective and stable representation and recognition results. The L1/2 and S1/2 regularizations are introduced to capture the sparsity of 1-D and 2-D, and to make the model have good statistical properties. Moreover, the cost-sensitive strategy is integrated into the model’s classification criteria to improve the imbalance of class distribution adaptively, especially for multiclass imbalanced data. A solver utilizing the mixed Gauss-Seidel and Jacobian ADMM algorithm is developed to obtain the approximate solution. Experiments on common public test databases show that the proposed model achieves competitive results compared with the latest published results and some deep-learning algorithms.

Cartesian coordinates scaffold stable spatial perception over time.

Visual systems exploit temporal continuity principles to achieve stable spatial perception, manifested as the serial dependence and central tendency effects. These effects are posited to reflect a smoothing process whereby past and present information integrates over time to decrease noise and stabilize perception. Meanwhile, the basic spatial coordinate—Cartesian versus polar—that scaffolds the integration process in two-dimensional continuous space remains unknown. The spatial coordinates are largely related to the allocentric and egocentric reference frames and presumably correspond with early and late processing stages in spatial perception. Here, four experiments consistently demonstrate that Cartesian outperforms polar coordinates in characterizing the serial bias—serial dependence and central tendency effect—in two-dimensional continuous spatial perception. The superiority of Cartesian coordinates is robust, independent of task environment (online and offline task), experimental length (short and long blocks), spatial context (shape of visual mask), and response modality (keyboard and mouse). Taken together, the visual system relies on the Cartesian coordinates for spatiotemporal integration to facilitate stable representation of external information, supporting the involvement of allocentric reference frame and top-down modulation in spatial perception over long time intervals.

Social Learning of a Spatial Task by Observation Alone.

Interactions between conspecifics are central to the acquisition of useful memories in the real world. Observational learning, i.e., learning a task by observing the success or failure of others, has been reported in many species, including rodents. However, previous work in rats with NMDA-receptor blockade has shown that even extensive observation of an unexplored space through a clear barrier is not sufficient to generate a stable hippocampal representation of that space. This raises the question of whether rats can learn a spatial task in a purely observed space from watching a conspecific, and if so, does this somehow stabilize their hippocampal representation? To address these questions, we designed an observational spatial task in a two-part environment that is nearly identical to that of the aforementioned electrophysiological study, in which an observer rat watches a demonstrator animal to learn the location of a hidden reward. Our results demonstrate that rats do not need to physically explore an environment to learn a reward location, provided a conspecific demonstrates where it is. We also show that the behavioral memory is not affected by NMDA receptor blockade, suggesting that the spatial representation underlying the behavior has been consolidated by observation alone.

A Novel Subspace-Aided Fault Detection Approach for the Drive Systems of Rolling Mills

This brief proposes a subspace-aided fault detection approach for the drive systems of strip rolling mills. Considering the impact of the unknown periodic load generated by the strip rolling process, the primary contributions are concluded as follows. First, this brief presents an approach to describe the subspace of the unknown/unmeasurable periodic load. Second, a fundamental frequency identification approach for the drive systems is proposed and then the subspace of the unknown periodic load can be constructed by the fundamental frequency. Third, this brief presents a subspace-aided fault detection approach to identify the data-driven stable kernel representation (SKR) of the closed-loop system by projecting the input–output (I/O) process data, so as to obtain a robust residual against the unknown periodic load. In addition, the effectiveness and performance of the approaches are verified by numerical examples and experimental data of the test rig for the drive systems of strip rolling mills. The results show that a robust residual generation against the unknown periodic load in the drive systems can be obtained and the robust subspace-aided fault detection can be achieved. Compared with the traditional method, the proposed approaches can improve the fault detection rate more effectively and be applied to the drive systems of strip rolling mills reliably.

Enhanced stability of complex sound representations relative to simple sounds in the auditory cortex.

Typical everyday sounds, such as those of speech or running water, are spectrotemporally complex. The ability to recognize complex sounds (CxSs) and their associated meaning is presumed to rely on their stable neural representations across time. The auditory cortex is critical for the processing of CxSs, yet little is known of the degree of stability of auditory cortical representations of CxSs across days. Previous studies have shown that the auditory cortex represents CxS identity with a substantial degree of invariance to basic sound attributes such as frequency. We therefore hypothesized that auditory cortical representations of CxSs are more stable across days than those of sounds that lack spectrotemporal structure such as pure tones (PTs). To test this hypothesis, we recorded responses of identified layer 2/3 auditory cortical excitatory neurons to both PTs and CxSs across days using two-photon calcium imaging in awake mice. Auditory cortical neurons showed significant daily changes of responses to both types of sounds, yet responses to CxSs exhibited significantly lower rates of daily change than those of PTs. Furthermore, daily changes in response profiles to PTs tended to be more stimulus-specific, reflecting changes in sound selectivity, compared with changes of CxS responses. Last, the enhanced stability of responses to CxSs was evident across longer time intervals as well. Together, these results suggest that spectrotemporally CxSs are more stably represented in the auditory cortex across time than PTs. These findings support a role of the auditory cortex in representing CxS identity across time.

MRRNet: Learning multiple region representation for video person re-identification

Video person re-identification is a crucial component of a robust surveillance system. Within a video clip, different human regions exhibit unique stability characteristics, which would be harmful to generating a discriminative representation. Unfortunately, prior works cannot effectively deal with the stability characteristics of different regions. To tackle this problem, we propose a Multiple Region Representation Network (MRRNet) that aims to discover the discriminative information from different human regions. Firstly, a Stable Region Representation (SRR) layer is proposed to capture important clues from the stable regions and exchange temporal information by cross-relation aware operation. Secondly, a Multiple Region Representation (MRR) layer is designed to address the unstable regions and preserve the attention on stable regions. Thirdly, SRR and MRR can be conveniently inserted into multiple stages of the deep residual networks and significantly improve the performance of the network. Comprehensive experiments validate the effectiveness of our network. Particularly, MRRNet achieves 86.7% mAP and 91.1% Rank-1 accuracy on the MARS dataset, which outperforms state-of-the-arts.

Semantic-Aware Representation Blending for Multi-Label Image Recognition with Partial Labels

Training the multi-label image recognition models with partial labels, in which merely some labels are known while others are unknown for each image, is a considerably challenging and practical task. To address this task, current algorithms mainly depend on pre-training classification or similarity models to generate pseudo labels for the unknown labels. However, these algorithms depend on sufficient multi-label annotations to train the models, leading to poor performance especially with low known label proportion. In this work, we propose to blend category-specific representation across different images to transfer information of known labels to complement unknown labels, which can get rid of pre-training models and thus does not depend on sufficient annotations. To this end, we design a unified semantic-aware representation blending (SARB) framework that exploits instance-level and prototype-level semantic representation to complement unknown labels by two complementary modules: 1) an instance-level representation blending (ILRB) module blends the representations of the known labels in an image to the representations of the unknown labels in another image to complement these unknown labels. 2) a prototype-level representation blending (PLRB) module learns more stable representation prototypes for each category and blends the representation of unknown labels with the prototypes of corresponding labels to complement these labels. Extensive experiments on the MS-COCO, Visual Genome, Pascal VOC 2007 datasets show that the proposed SARB framework obtains superior performance over current leading competitors on all known label proportion settings, i.e., with the mAP improvement of 4.6%, 4.6%, 2.2% on these three datasets when the known label proportion is 10%. Codes are available at https://github.com/HCPLab-SYSU/HCP-MLR-PL.

Hybrid quadrature moment method for accurate and stable representation of non-Gaussian processes applied to bubble dynamics.

Solving the population balance equation (PBE) for the dynamics of a dispersed phase coupled to a continuous fluid is expensive. Still, one can reduce the cost by representing the evolving particle density function in terms of its moments. In particular, quadrature-based moment methods (QBMMs) invert these moments with a quadrature rule, approximating the required statistics. QBMMs have been shown to accurately model sprays and soot with a relatively compact set of moments. However, significantly non-Gaussian processes such as bubble dynamics lead to numerical instabilities when extending their moment sets accordingly. We solve this problem by training a recurrent neural network (RNN) that adjusts the QBMM quadrature to evaluate unclosed moments with higher accuracy. The proposed method is tested on a simple model of bubbles oscillating in response to a temporally fluctuating pressure field. The approach decreases model-form error by a factor of 10 when compared with traditional QBMMs. It is both numerically stable and computationally efficient since it does not expand the baseline moment set. Additional quadrature points are also assessed, optimally placed and weighted according to an additional RNN. These points further decrease the error at low cost since the moment set is again unchanged. This article is part of the theme issue 'Data-driven prediction in dynamical systems'.