Similarity In Representations Research Articles

Early life adversity is associated with greater similarity in neural representations of ambiguous and threatening stimuli.

Exposure to early life adversity (ELA) is hypothesized to sensitize threat-responsive neural circuitry. This may lead individuals to overestimate threat in the face of ambiguity, a cognitive-behavioral phenotype linked to poor mental health. The tendency to process ambiguity as threatening may stem from difficulty distinguishing between ambiguous and threatening stimuli. However, it is unknown how exposure to ELA relates to neural representations of ambiguous and threatening stimuli, or how processing of ambiguity following ELA relates to psychosocial functioning. The current fMRI study examined multivariate representations of threatening and ambiguous social cues in 41 emerging adults (aged 18 to 19 years). Using representational similarity analysis, we assessed neural representations of ambiguous and threatening images within affective neural circuitry and tested whether similarity in these representations varied by ELA exposure. Greater exposure to ELA was associated with greater similarity in neural representations of ambiguous and threatening images. Moreover, individual differences in processing ambiguity related to global functioning, an association that varied as a function of ELA. By evidencing reduced neural differentiation between ambiguous and threatening cues in ELA-exposed emerging adults and linking behavioral responses to ambiguity to psychosocial wellbeing, these findings have important implications for future intervention work in at-risk, ELA-exposed populations.

Intersubject similarity in neural representations underlies shared episodic memory content

Individuals generally form their unique memories from shared experiences, yet the neural representational mechanisms underlying this subjectiveness of memory are poorly understood. The current study addressed this important question from the cross-subject neural representational perspective, leveraging a large functional magnetic resonance imaging dataset (n = 415) of a face-name associative memory task. We found that individuals' memory abilities were predicted by their synchronization to the group-averaged, canonical trial-by-trial activation level and, to a lesser degree, by their similarity to the group-averaged representational patterns during encoding. More importantly, the memory content shared between pairs of participants could be predicted by their shared local neural activation pattern, particularly in the angular gyrus and ventromedial prefrontal cortex, even after controlling for differences in memory abilities. These results uncover neural representational mechanisms for individualized memory and underscore the constructive nature of episodic memory.

Contrastive Learning of Subject-Invariant EEG Representations for Cross-Subject Emotion Recognition

EEG signals have been reported to be informative and reliable for emotion recognition in recent years. However, the inter-subject variability of emotion-related EEG signals still poses a great challenge for the practical applications of EEG-based emotion recognition. Inspired by recent neuroscience studies on inter-subject correlation, we proposed a Contrastive Learning method for Inter-Subject Alignment (CLISA) to tackle the cross-subject emotion recognition problem. Contrastive learning was employed to minimize the inter-subject differences by maximizing the similarity in EEG signal representations across subjects when they received the same emotional stimuli in contrast to different ones. Specifically, a convolutional neural network was applied to learn inter-subject aligned spatiotemporal representations from EEG time series in contrastive learning. The aligned representations were subsequently used to extract differential entropy features for emotion classification. CLISA achieved state-of-the-art cross-subject emotion recognition performance on our THU-EP dataset with 80 subjects and the publicly available SEED dataset with 15 subjects. It could generalize to unseen subjects or unseen emotional stimuli in testing. Furthermore, the spatiotemporal representations learned by CLISA could provide insights into the neural mechanisms of human emotion processing.

Neural representations of ambiguous affective stimuli and resilience to anxiety in emerging adults

The tendency to interpret ambiguous stimuli as threatening has been associated with a range of anxiety disorders. Responses to ambiguity may be particularly relevant to mental health during the transition from adolescence to adulthood (“emerging adulthood”), when individuals encounter unfamiliar challenges and navigate novel social situations. However, it remains unclear whether neural representations of ambiguity relate to risk for anxiety. The present study sought to examine whether multivariate representations of ambiguity — and their similarity to representations of threat — relate to appraisals of ambiguity or anxiety in a sample of emerging adults. Participants (N = 41) viewed threatening (angry), nonthreatening (happy), and ambiguous (surprised) facial stimuli while undergoing fMRI. Outside of the scanner, participants were presented with the same stimuli and categorized the ambiguous faces as positive or negative. Using representational similarity analyses (RSA), we investigated whether the degree of pattern similarity in responses to ambiguous, nonthreatening, and threatening faces within the amygdala related to appraisals of ambiguous stimuli and anxiety symptomatology. We found that individuals who evidenced greater similarity (i.e., less differentiation) in neural representations of ambiguous and nonthreatening faces within the left amygdala reported lower concurrent anxiety. Additionally, trial-level pattern similarity predicted subsequent appraisals of ambiguous stimuli. These findings provide insight into how neural representations of ambiguity relate to risk or resilience for the development of anxiety.

Shared neural representations and temporal segmentation of political content predict ideological similarity.

Despite receiving the same sensory input, opposing partisans often interpret political content in disparate ways. Jointly analyzing controlled and naturalistic functional magnetic resonance imaging data, we uncover the neurobiological mechanisms explaining how these divergent political viewpoints arise. Individuals who share an ideology have more similar neural representations of political words, experience greater neural synchrony during naturalistic political content, and temporally segment real-world information into the same meaningful units. In the striatum and amygdala, increasing intersubject similarity in neural representations of political concepts during a word reading task predicts enhanced synchronization of blood oxygen level-dependent time courses when viewing real-time, inflammatory political videos, revealing that polarization can arise from differences in the brain's affective valuations of political concepts. Together, this research shows that political ideology is shaped by semantic representations of political concepts processed in an environment free of any polarizing agenda and that these representations bias how real-world political information is construed into a polarized perspective.

Element similarity in high-dimensional materials representations

Elements can be represented as vectors in a high-dimensional chemical space. We explore the distance and correlation between these vectors for different machine learning models.

Do all mice smell the same? Chemosensory cues from inbred and wild mouse strains elicit stereotypic sensory representations in the accessory olfactory bulb

BackgroundFor many animals, chemosensory cues are vital for social and defensive interactions and are primarily detected and processed by the vomeronasal system (VNS). These cues are often inherently associated with ethological meaning, leading to stereotyped behaviors. Thus, one would expect consistent representation of these stimuli across different individuals. However, individuals may express different arrays of vomeronasal sensory receptors and may vary in the pattern of connections between those receptors and projection neurons in the accessory olfactory bulb (AOB). In the first part of this study, we address the ability of individuals to form consistent representations despite these potential sources of variability. The second part of our study is motivated by the fact that the majority of research on VNS physiology involves the use of stimuli derived from inbred animals. Yet, it is unclear whether neuronal representations of inbred-derived stimuli are similar to those of more ethologically relevant wild-derived stimuli.ResultsFirst, we compared sensory representations to inbred, wild-derived, and wild urine stimuli in the AOBs of males from two distinct inbred strains, using them as proxies for individuals. We found a remarkable similarity in stimulus representations across the two strains. Next, we compared AOB neuronal responses to inbred, wild-derived, and wild stimuli, again using male inbred mice as subjects. Employing various measures of neuronal activity, we show that wild-derived and wild stimuli elicit responses that are broadly similar to those from inbred stimuli: they are not considerably stronger or weaker, they show similar levels of sexual dimorphism, and when examining population-level activity, cluster with inbred mouse stimuli.ConclusionsDespite strain-specific differences and apparently random connectivity, the AOB can maintain stereotypic sensory representations for broad stimulus categories, providing a substrate for common stereotypical behaviors. In addition, despite many generations of inbreeding, AOB representations capture the key ethological features (i.e., species and sex) of wild-derived and wild counterparts. Beyond these broad similarities, representations of stimuli from wild mice are nevertheless distinct from those elicited by inbred mouse stimuli, suggesting that laboratory inbreeding has indeed resulted in marked modifications of urinary secretions.

The effects of word concreteness on cross-language neural pattern similarity during semantic categorization

Previous behavioral studies have suggested that word concreteness affects the similarity of semantic representations between native and second languages. However, it is unclear whether and how word concreteness influences the cross-language similarity in neural representations. In the present study, we quantified the neural pattern similarity between native and second languages by using representational similarity analysis, and examined the modulatory effect of word concreteness on cross-language similarity in neural representations. Twenty-six Chinese-English bilinguals performed a semantic decision task during fMRI scan. Results showed that concrete and abstract words elicited stronger activation in the right sensorimotor cortex for nonverbal processing and the left prefrontal cortex for verbal processing, respectively. Consistently, representational similarity analysis revealed that concrete and abstract words showed higher cross-language neural pattern similarity in those brain regions for nonverbal and verbal processing, respectively. More importantly, we found that the left middle temporal gyrus contained item-specific cross-language semantic representations for abstract words. These results suggest that word concreteness plays an important role in influencing cross-language similarity in neural representations. • RSA is used to quantify cross-language pattern similarity. • Concrete and abstract words elicit distinct neural activity patterns. • LMTG shows item-specific cross-language pattern similarity for abstract words.

Natural scene representations in the gamma band are prototypical across subjects

Prototypical brain responses describe similarity in neural representations between subjects in response to a natural stimulus. During natural movie viewing, for example, inter-subject correlation (ISC) measured by fMRI is high in visual areas (Hasson et al., 2004). But the electrophysiological basis for this fMRI ISC has been controversial. Previous reports have only found ISC in low frequency bands—below 12 ​Hz (Chang et al., 2015). These findings stand in contrast to reports that gamma band oscillations—30 to 90 ​Hz—are highly stimulus-driven in visual cortex (Perry et al., 2015). To resolve this discrepancy, we carried out both ISC estimation and a novel inter-subject representational correlation analysis across six frequency bands extracted from MEG data of 24 subjects who each viewed four 5-min clips of an underwater documentary. Region-of-interest-based and vertex-based temporal ISC estimates confirmed that low-frequency bands are significantly synchronized in visual areas and that gamma band has low temporal correlation. We also found the representational geometry of movie scenes were related to structural statistics from the stimuli. Crucially, our results show that the gamma band oscillations also reflect prototypical brain response in scene representations formed in response to naturalistic stimuli as revealed by inter-subject representational correlation.

How meaningful are similarities in deep trajectory representations?

Finding similar trajectories is an important task in moving object databases. However, classical similarity models face several limitations, including scalability and robustness. Recently, an approach named t2vec proposed transforming trajectories into points in a high dimensional vector space, and this transformation approximately keeps distances between trajectories. t2vec overcomes that scalability limitation: Now it is possible to cluster millions of trajectories. However, the semantics of the learned similarity values – and whether they are meaningful – is an open issue. One can ask: How does the configuration of t2vec affect the similarity values of trajectories? Is the notion of similarity in t2vec similar, different, or even superior to existing models? As for any neural-network-based approach, inspecting the network does not help to answer these questions. So the problem we address in this paper is how to assess the meaningfulness of similarity in deep trajectory representations. Our solution is a methodology based on a set of well-defined, systematic experiments. We compare t2vec to classical models in terms of robustness and their semantics of similarity, using two real-world datasets. We give recommendations which model to use in possible application scenarios and use cases. We conclude that using t2vec in combination with classical models may be the best way to identify similar trajectories. Finally, to foster scientific advancement, we give the public access to all trained t2vec models and experiment scripts. To our knowledge, this is the biggest collection of its kind.

Mosaic representations of odors in the input and output layers of the mouse olfactory bulb.

The elementary stimulus features encoded by the olfactory system remain poorly understood. We examined the relationship between 1,666 physical-chemical descriptors of odors and the activity of olfactory bulb inputs and outputs in awake mice. Glomerular and mitral and tufted cell responses were sparse and locally heterogeneous, with only a weak dependence of their positions on physical-chemical properties. Odor features represented by ensembles of mitral and tufted cells were overlapping but distinct from those represented in glomeruli, which is consistent with an extensive interplay between feedforward and feedback inputs to the bulb. This reformatting was well described as a rotation in odor space. The physical-chemical descriptors accounted for a small fraction in response variance, and the similarity of odors in the physical-chemical space was a poor predictor of similarity in neuronal representations. Our results suggest that commonly used physical-chemical properties are not systematically represented in bulbar activity and encourage further searches for better descriptors of odor space.

What Do Language Representations Really Represent?

A neural language model trained on a text corpus can be used to induce distributed representations of words, such that similar words end up with similar representations. If the corpus is multilingual, the same model can be used to learn distributed representations of languages, such that similar languages end up with similar representations. We show that this holds even when the multilingual corpus has been translated into English, by picking up the faint signal left by the source languages. However, just as it is a thorny problem to separate semantic from syntactic similarity in word representations, it is not obvious what type of similarity is captured by language representations. We investigate correlations and causal relationships between language representations learned from translations on one hand, and genetic, geographical, and several levels of structural similarity between languages on the other. Of these, structural similarity is found to correlate most strongly with language representation similarity, whereas genetic relationships—a convenient benchmark used for evaluation in previous work—appears to be a confounding factor. Apart from implications about translation effects, we see this more generally as a case where NLP and linguistic typology can interact and benefit one another.

A mathematical theory of semantic development in deep neural networks

An extensive body of empirical research has revealed remarkable regularities in the acquisition, organization, deployment, and neural representation of human semantic knowledge, thereby raising a fundamental conceptual question: What are the theoretical principles governing the ability of neural networks to acquire, organize, and deploy abstract knowledge by integrating across many individual experiences? We address this question by mathematically analyzing the nonlinear dynamics of learning in deep linear networks. We find exact solutions to this learning dynamics that yield a conceptual explanation for the prevalence of many disparate phenomena in semantic cognition, including the hierarchical differentiation of concepts through rapid developmental transitions, the ubiquity of semantic illusions between such transitions, the emergence of item typicality and category coherence as factors controlling the speed of semantic processing, changing patterns of inductive projection over development, and the conservation of semantic similarity in neural representations across species. Thus, surprisingly, our simple neural model qualitatively recapitulates many diverse regularities underlying semantic development, while providing analytic insight into how the statistical structure of an environment can interact with nonlinear deep-learning dynamics to give rise to these regularities.

The effects of enactment and intention accessibility on prospective memory performance.

The intention-superiority effect denotes faster response latencies to stimuli linked with a prospective memory task compared to stimuli linked with no prospective task or with a cancelled task. It is generally assumed that the increased accessibility of intention-related materials contributes to successful execution of prospective memory tasks at an appropriate opportunity. In two experiments we investigated the relationship between the intention-superiority effect and actual prospective memory performance under relatively realistic conditions. We also manipulated enactment versus observation encoding to further investigate the similarity in representations of enacted and to-be-enacted tasks. Additionally, Experiment 1 included a control condition to investigate the development of the intention-superiority effect over time. Participants were asked to perform prospective tasks at the end of the experiment to prepare the room for the next participant. They studied these preparatory tasks at the beginning of the experiment either by enacting them themselves or by observing the experimenter perform them. In Experiment 2, participants in a control condition did not intend to perform prospective tasks. We observed a smaller intention-superiority effect after enactment encoding than after observation encoding, but only if response latencies were assessed immediately before the prospective memory task. In addition, Experiment 2 suggested that the size of the intention-superiority effect is related to successful prospective memory performance, thus providing evidence for a functional relationship between accessibility and memory.

Cohesion, Similarity and Value in Parent-Child Representations of Albanian and Serbian Immigrant and Italian Native Children

Cohesion, Similarity and Value in Parent-Child Representations of Albanian and Serbian Immigrant and Italian Native Children

Tourist Destination Image in the New Technological Environments: Dialogism under the Public Relations Gaze

The dependence and influence of new media on social and organizational life have led to a new orientation of the communicational practices of organizations and to a research paradigm shift to a cocriational model within public relations theory. This model is based on dialogic rhetoric and dialogue that requires a joint production of content and meanings by organizations and publics, which have a central role in shaping image. The new social and communicational paradigms have also an impact on tourism research, implying the need to understand the role of new technologies, particularly websites, in image formation. The purpose of this study is to investigate the role of website in the formation of a tourist destination image (Portugal) through the understanding of the degree of adoption of dialogical principles on websites by official and unofficial tourist organizations and the identifying of denotative and connotative elements that are represented on published and shared photographs in these websites. The research corpus consisted of websites of entities and photographs published and shared on these websites. The methodology used is quantitative and qualitative, based on content analysis of data of websites and photographs. Results indicate that official tourist entities adopted the dialogic principles in a higher degree than unofficials. There is also a higher adoption of the principles related to technical dimension than dialogical dimension. Results confirm a relation between the adoption of dialogical link and an increased openness by entities to the possibility of photo sharing. The analysis of denotative and connotative elements of photography reflects a similarity in representations of entities and tourists. This study contributes to both theoretical perspectives, because it provides a conceptual scheme of destination image formation online and to a practical perspective, because it also suggests guidelines of dialogical practices to incorporate in websites of tourism entities.

Word recognition in deaf readers: Cross-language activation of German Sign Language and German

ABSTRACTThis study addressed visual word recognition in deaf bilinguals who are proficient in German Sign Language (DGS) and German. The study specifically investigated whether DGS signs are activated during a monolingual German word recognition task despite the lack of similarity in German orthographic representations and DGS phonological representations. Deaf DGS–German bilinguals saw pairs of German words and decided whether the words were semantically related. Half of the experimental items had phonologically related translation equivalents in DGS. Participants were slower to reject semantically unrelated word pairs when the translation equivalents were phonologically related in DGS than when the DGS translations were phonologically unrelated. However, this was not the case in Turkish–German hearing bilinguals who do not have sign language knowledge. The results indicate that lexical representations are associated cross-linguistically in the bilingual lexicon irrespective of their orthographic or phonological form. Implications of these results for reading development in deaf German bilinguals are discussed.

Amodal completion is modulated by lightness similarity

The strength of amodal completion is known to be modulated by contour relationships and global shape. Some researchers have shown that amodal completion also depends on surface similarity, but they have not distinguished the relative importance of similarity in surface representations either pre or post lightness constancy. In the experiments reported here, we aimed to determine whether amodal completion depends on processes that occur either before or after the establishment of lightness constancy. We used computer rendering techniques to vary the consistency of a cast shadow with a decrement in luminance on one side of a partially occluded surface. We found that perceived completion depended on the consistency of the decrement in surface luminance with the orientation of a cast shadow (Exp. 1). In Experiment 2, we generated occluded surface fragments that could be either luminance-matched or lightness-matched to surface fragments on the opposite side of the occluder, and we found that the strength of amodal completion depended primarily on similarity in the perceived surface reflectance. In Experiments 3 and 4, we performed apparent-motion tasks to obtain converging evidence for grouping on the basis of perceived similarity in lightness. We found that matching surfaces in lightness significantly improved the apparent motion of surfaces behind an occluder, as compared with surfaces that were matched in contrast alone. These results suggest that amodal completion depends on the perceived similarity in lightness of partially occluded surface fragments.

Structure-preserving model reduction of passive and quasi-active neurons

The spatial component of input signals often carries information crucial to a neuron's function, but models mapping synaptic inputs to the transmembrane potential can be computationally expensive. Existing reduced models of the neuron either merge compartments, thereby sacrificing the spatial specificity of inputs, or apply model reduction techniques that sacrifice the underlying electrophysiology of the model. We use Krylov subspace projection methods to construct reduced models of passive and quasi-active neurons that preserve both the spatial specificity of inputs and the electrophysiological interpretation as an RC and RLC circuit, respectively. Each reduced model accurately computes the potential at the spike initiation zone (SIZ) given a much smaller dimension and simulation time, as we show numerically and theoretically. The structure is preserved through the similarity in the circuit representations, for which we provide circuit diagrams and mathematical expressions for the circuit elements. Furthermore, the transformation from the full to the reduced system is straightforward and depends on intrinsic properties of the dendrite. As each reduced model is accurate and has a clear electrophysiological interpretation, the reduced models can be used not only to simulate morphologically accurate neurons but also to examine computations performed in dendrites.

Spatial language and spatial representation: a cross-linguistic comparison

We examined the relationship between spatial language and spatial memory by comparing native English, Japanese, and Korean speakers' naming of spatial locations and their spatial memory for the same set of locations. We focused on two kinds of spatial organization: axial structure of the reference object, and contact/support with respect to its surface. The results of two language (naming) tasks showed similar organization across the three language groups in terms of axial structure, but differences in organization in terms of contact/support. In contrast, the results of two memory tasks were the same across language groups for both axial structure and contact/support. Moreover, the relationship between spatial language and spatial memory in the two sets of tasks did not show a straightforward isomorphism between the two systems. We conclude that spatial language and spatial memory engage the same kinds of spatial properties, suggesting similarity in the foundations of the two systems. However, the two systems appear to be partially independent: the preservation of particular spatial properties was not mandatory across languages, nor across memory tasks, and cross-linguistic differences in spatial language did not lead to differences in the non-linguistic encoding of location. We speculate that the similarity in linguistic and non-linguistic representations of space may emerge as a functional consequence of negotiating the spatial world.