Confidence-Controlled Hebbian Learning Efficiently Extracts Category Membership From Stimuli Encoded in View of a Categorization Task

In experiments on perceptual decision making, individuals learn a categorization task through trial-and-error protocols. We explore the capacity of a decision-making attractor network to learn a categorization task through reward-based, Hebbian-type modifications of the weights incoming from the stimulus encoding layer. For the latter, we assume a standard layer of a large number of stimu lus-specific neurons. Within the general framework of Hebbian learning, we have hypothesized that the learning rate is modulated by the reward at each trial. Surprisingly, we find that when the coding layer has been optimized in view of the categorization task, such reward-modulated Hebbian learning (RMHL) fails to extract efficiently the category membership. In previous work, we showed that the attractor neural networks' nonlinear dynamics accounts for behavioral confidence in sequences of decision trials. Taking advantage of these findings, we propose that learning is controlled by confidence, as computed from the neural activity of the decision-making attractor network. Here we show that this confidence-controlled, reward-based Hebbian learning efficiently extracts categorical information from the optimized coding layer. The proposed learning rule is local and, in contrast to RMHL, does not require storing the average rewards obtained on previous trials. In addition, we find that the confidence-controlled learning rule achieves near-optimal performance. In accordance with this result, we show that the learning rule approximates a gradient descent method on a maximizing reward cost function.

Older and Younger Adults Perform Similarly in an Iterated Trust Game

In social contexts, aging is typically associated with a greater reliance on heuristics, such as categorical information and stereotypes. The present research examines younger and older adults’ use of individuating and age-based categorical information when gauging whether or not to trust unfamiliar targets. In an adaptation of the iterated Trust Game, participants had to predict the cooperative tendencies of their partners to earn economic rewards in first encounters – in a context in which they knew nothing about their partners, and across repeated interactions – in a context in which they could learn the individual cooperative tendency of each partner. In line with previous research, we expected all participants to rely on stereotypes in first encounters, and progressively learn to disregard stereotypes to focus on individuating behavioral cues across repeated interactions. Moreover, we expected older participants to rely more on social categories than younger participants. Our results indicate that overall, both the elderly and the young adopted an individuating approach to predict the cooperative behaviors of their partners across trials. However, older adults more consistently relied on gender (but not age) stereotypes to make cooperation decisions at zero acquaintance. The impact of context, motivation, and relevance of categorical information in impression formation is discussed.

Similarity Search and Correlation-Based Exploratory Analysis in EHRs: A Case Study with COVID-19 Databases

With the COVID-19 pandemic, many hospitals have collected Electronic Health Records (EHRs) from patients and shared them publicly. EHRs include heterogeneous attribute types, such as image exams, numerical, textual, and categorical information. Simply posing similarity queries over EHRs can underestimate the semantics and potential information of particular attributes and thus would be best supported by exploratory data analysis methods. Thus, we propose the Sketch method for comparing EHRs by similarity to provide a tool for a correlation-based exploratory analysis over different attributes. Sketch computes the overall data correlation considering the distance space of every attribute. Further, it employs both ANOVA and association rules with lift correlations to study the relationship between variables, allowing a deep data analysis. As a case study, we employed two open databases of COVID-19 cases, showing that specialists can benefit from the inference modules of Sketch to analyze EHRs. Sketch found strong correlations among tuples and attributes, with statistically significant results. The exploratory analysis has shown to complement the similarity search task, identifying and evaluating patterns discovered from heterogeneous attributes.

A fingerprint of a heterogeneous data set

In this paper, we describe the fingerprint method, a technique to classify bags of mixed-type measurements. The method was designed to solve a real-world industrial problem: classifying industrial plants (individuals at a higher level of organisation) starting from the measurements collected from their production lines (individuals at a lower level of organisation). In this specific application, the categorical information attached to the numerical measurements induced simple mixture-like structures on the global multivariate distributions associated with different classes. The fingerprint method is designed to compare the mixture components of a given test bag with the corresponding mixture components associated with the different classes, identifying the most similar generating distribution. When compared to other classification algorithms applied to several synthetic data sets and the original industrial data set, the proposed classifier showed remarkable improvements in performance.

A continuum of invariant sensory and behavioral-context perceptual coding in secondary somatosensory cortex

A crucial role of cortical networks is the conversion of sensory inputs into perception. In the cortical somatosensory network, neurons of the primary somatosensory cortex (S1) show invariant sensory responses, while frontal lobe neuronal activity correlates with the animal’s perceptual behavior. Here, we report that in the secondary somatosensory cortex (S2), neurons with invariant sensory responses coexist with neurons whose responses correlate with perceptual behavior. Importantly, the vast majority of the neurons fall along a continuum of combined sensory and categorical dynamics. Furthermore, during a non-demanding control task, the sensory responses remain unaltered while the sensory information exhibits an increase. However, perceptual responses and the associated categorical information decrease, implicating a task context-dependent processing mechanism. Conclusively, S2 neurons exhibit intriguing dynamics that are intermediate between those of S1 and frontal lobe. Our results contribute relevant evidence about the role that S2 plays in the conversion of touch into perception.

Conceptual distinctiveness does not improve visual working memory for objects but aids object-location binding

Visual working memory (VWM) is prone to interference from stored items competing for its limited capacity. These competitive interactions can arise from different sources. For example, one such source is poor item distinctiveness causing a failure to discriminate between items sharing common features. Another source of interference is imperfect binding, a problem of determining which of the remembered features belonged to which object or which item was in which location. In two experiments, we studied how the conceptual distinctiveness of real-world objects (i.e., whether the objects belong to the same or different basic categories) affects VWM for objects and object-location binding. In Experiment 1, we found that distinctiveness did not affect memory for object identities or for locations, but low-distinctive objects were more frequently reported at “swapped” locations that originally went with different objects. In Experiment 2 we found evidence that the effect of distinctiveness on the object-location swaps was due to the use of categorical information for binding. In particular, we found that observers swapped the location of a tested object with another object from the same category more frequently than with any of the objects from another category. This suggests that observers can use some coarse category-location information when objects are conceptually distinct. Taken together, our findings suggest that object distinction and object-location binding act upon different components of VWM.

Nonverbal category knowledge limits the amount of information encoded in object representations: EEG evidence from 12-month-old infants

To what extent does language shape how we think about the world? Studies suggest that linguistic symbols expressing conceptual categories (‘apple’, ‘squirrel’) make us focus on categorical information (e.g. that you saw a squirrel) and disregard individual information (e.g. whether that squirrel had a long or short tail). Across two experiments with preverbal infants, we demonstrated that it is not language but nonverbal category knowledge that determines what information is packed into object representations. Twelve-month-olds ( N = 48) participated in an electroencephalography (EEG) change-detection task involving objects undergoing a brief occlusion. When viewing objects from unfamiliar categories, infants detected both across- and within-category changes, as evidenced by their negative central wave (Nc) event-related potential. Conversely, when viewing objects from familiar categories, they did not respond to within-category changes, which indicates that nonverbal category knowledge interfered with the representation of individual surface features necessary to detect such changes. Furthermore, distinct patterns of γ and α oscillations between familiar and unfamiliar categories were evident before and during occlusion, suggesting that categorization had an influence on the format of recruited object representations. Thus, we show that nonverbal category knowledge has rapid and enduring effects on object representation and discuss their functional significance for generic knowledge acquisition in the absence of language.

The Memory for Latent Representations: An Account of Working Memory that Builds on Visual Knowledge for Efficient and Detailed Visual Representations

Visual knowledge obtained from our lifelong experience of the world plays a critical role in our ability to build short-term memories. We propose a mechanistic explanation of how working memories are built from the latent representations of visual knowledge and can then be reconstructed. The proposed model, Memory for Latent Representations (MLR), features a variational autoencoder with an architecture that corresponds broadly to the human visual system and an activation-based binding pool of neurons that binds items' attributes to tokenized representations. The simulation results revealed that shape information for stimuli that the model was trained on, can be encoded and retrieved efficiently from latents in higher levels of the visual hierarchy. On the other hand, novel patterns that are completely outside the training set can be stored from a single exposure using only latents from early layers of the visual system. Moreover, a given stimulus in working memory can have multiple codes, representing specific visual features such as shape or color, in addition to categorical information. Finally, we validated our model by testing a series of predictions against behavioral results acquired from WM tasks. The model provides a compelling demonstration of visual knowledge yielding the formation of compact visual representation for efficient memory encoding.