Task Structure Research Articles

Progress and opportunities of foundation models in bioinformatics.

Bioinformatics has undergone a paradigm shift in artificial intelligence (AI), particularly through foundation models (FMs), which address longstanding challenges in bioinformatics such as limited annotated data and data noise. These AI techniques have demonstrated remarkable efficacy across various downstream validation tasks, effectively representing diverse biological entities and heralding a new era in computational biology. The primary goal of this survey is to conduct a general investigation and summary of FMs in bioinformatics, tracing their evolutionary trajectory, current research landscape, and methodological frameworks. Our primary focus is on elucidating the application of FMs to specific biological problems, offering insights to guide the research community in choosing appropriate FMs for tasks like sequence analysis, structure prediction, and function annotation. Each section delves into the intricacies of the targeted challenges, contrasting the architectures and advancements of FMs with conventional methods and showcasing their utility across different biological domains. Further, this review scrutinizes the hurdles and constraints encountered by FMs in biology, including issues of data noise, model interpretability, and potential biases. This analysis provides a theoretical groundwork for understanding the circumstances under which certain FMs may exhibit suboptimal performance. Lastly, we outline prospective pathways and methodologies for the future development of FMs in biological research, facilitating ongoing innovation in the field. This comprehensive examination not only serves as an academic reference but also as a roadmap for forthcoming explorations and applications of FMs in biology.

From task-general towards task-specific cognitive operations in a few minutes? Working memory performance as an adaptive process.

Measurement of cognitive functions is typically based on the implicit assumption that the mental architecture underlying cognitive task performance is constant throughout the task. In contrast, skill learning theory implies that cognitively demanding task performance is an adaptive process that progresses from initial heavy engagement of effortful and task-general metacognitive and executive control processes towards more automatic and task-specific performance. However, this hypothesis is rarely applied to the short time spans of traditional cognitive tasks such as working memory (WM) tasks. We utilised longitudinal structural equation models on two well-powered data sets to test the hypothesis that the initial stages of WM task performances load heavily on a task-general g-factor and then start to diverge towards factors specific to task structure. In line with the hypothesis, data from the first experiment (N = 296) were successfully fitted in a model with task-initial unity of the WM paradigm-specific latent factors, after which their intercorrelations started to diverge. The second experiment (N = 201) replicated this pattern except for one paradigm-specific latent factor. These preliminary results suggest that the processes underlying WM task performance tend to progress rapidly from more task-general towards task-specific, in line with the cognitive skill learning framework. Such task-internal dynamics has important implications for the measurement of complex cognitive functions.

Flexible Work in the Public Sector: A Dual Perspective on Cognitive Benefits and Costs in Remote Work Environments

The COVID-19 pandemic has accelerated the adoption of flexible work arrangements within the Italian public administration. While much of the existing research has focused on the drawbacks of such arrangements, there has been less exploration of their benefits. Cognitive demands related to the structure of work activities, planning of working hours, planning of workplaces, and coordination with others, under flexible working conditions, might be considered as job resources that act as challenging demands within the Job Demand-Resource (JD-R) model. This study aimed to explore how the “cognitive challenge of flexible work” (CCFW) impact job satisfaction through home-based performance, taking into account the role of weekly working hours on home-based performance. Furthermore, the potential moderating role of cognitive and physical job demands between CCFW and home-based performance was explored. Using structural equation modeling on data from 484 public employees, the findings confirmed the positive impact of the structure of work tasks and planning of working times on both job satisfaction and home-based performance. In addition, cognitive demands (i.e., perception of cognitive work overload) played a moderating role in the mediated relationship between coordinating with others on job satisfaction and the structure of working tasks on job satisfaction through home-based performance.

A Multi-Task Convolutional Neural Network Relative Radiometric Calibration Based on Temporal Information

Due to the continuous degradation of onboard satellite instruments over time, satellite images undergo degradation, necessitating calibration for tasks reliant on satellite data. The previous relative radiometric calibration methods are mainly categorized into traditional methods and deep learning methods. The traditional methods involve complex computations for each calibration, while deep-learning-based approaches tend to oversimplify the calibration process, utilizing generic computer vision models without tailored structures for calibration tasks. In this paper, we address the unique challenges of calibration by introducing a novel approach: a multi-task convolutional neural network calibration model leveraging temporal information. This pioneering method is the first to integrate temporal dynamics into the architecture of neural network calibration models. Extensive experiments conducted on the FY3A/B/C VIRR datasets showcase the superior performance of our approach compared to the existing state-of-the-art traditional and deep learning methods. Furthermore, tests with various backbones confirm the broad applicability of our framework across different convolutional neural networks.

A line balancing problem with parallel workers and cycle time minimization

We study the problem of balancing assembly lines with parallel workers, motivated by features observed at a third-party logistics provider. The assembly line differs from the two known types of line balancing problems in the literature since it determines simultaneously the line cycle time and the number of workers per station. Furthermore, each station can be occupied by more than one worker and a restriction on the minimum number of stations is imposed. This new type of line balancing problem minimizes the line cycle time, where two types of decisions must be made: how to group the tasks into stations and how many workers to assign to each station. We adapt mathematical models based on assembly line balancing problems with parallel stations. Furthermore, we propose a new model based on network flow formulations for the special case of lines with a serial task structure, frequently observed at the company. We perform an extensive computational study with realistic instances in order to compare the speed of solving different formulations, which is important as such a problem is at the operational level. We also provide a sensitivity analysis of key parameters aiming to better understand the trade-offs and provide meaningful managerial insights.

Functional benefits of continuous vs. categorical listening strategies on the neural encoding and perception of noise-degraded speech

Acoustic information in speech changes continuously, yet listeners form discrete perceptual categories to ease the demands of perception. Being a more continuous/gradient as opposed to a more discrete/categorical listener may be further advantageous for understanding speech in noise by increasing perceptual flexibility and resolving ambiguity. The degree to which a listener’s responses to a continuum of speech sounds are categorical versus continuous can be quantified using visual analog scaling (VAS) during speech labeling tasks. Here, we recorded event-related brain potentials (ERPs) to vowels along an acoustic–phonetic continuum (/u/ to /a/) while listeners categorized phonemes in both clean and noise conditions. Behavior was assessed using standard two alternative forced choice (2AFC) and VAS paradigms to evaluate categorization under task structures that promote discrete vs. continuous hearing, respectively. Behaviorally, identification curves were steeper under 2AFC vs. VAS categorization but were relatively immune to noise, suggesting robust access to abstract, phonetic categories even under signal degradation. Behavioral slopes were correlated with listeners’ QuickSIN scores; shallower slopes corresponded with better speech in noise performance, suggesting a perceptual advantage to noise degraded speech comprehension conferred by a more gradient listening strategy. At the neural level, P2 amplitudes and latencies of the ERPs were modulated by task and noise; VAS responses were larger and showed greater noise-related latency delays than 2AFC responses. More gradient responders had smaller shifts in ERP latency with noise, suggesting their neural encoding of speech was more resilient to noise degradation. Interestingly, source-resolved ERPs showed that more gradient listening was also correlated with stronger neural responses in left superior temporal gyrus. Our results demonstrate that listening strategy modulates the categorical organization of speech and behavioral success, with more continuous/gradient listening being advantageous to sentential speech in noise perception.

Blockchain-driven decentralized identity management: An interdisciplinary review and research agenda

The rise of blockchain technology has sparked interest in decentralized identity management (DIdM). However, DIdM's interdisciplinary nature has led to a fragmented understanding. We propose a “Task Structure-Technological Properties-Fit” framework to clarify the application of DIdM across tasks and contexts. We conducted a comprehensive review of 149 DIdM papers to define task structure (task goals and stakeholder values), task goals (identity value creation and value maintenance), stakeholders (users, service providers, identity issuers, regulators, and infrastructure builders), and DIdM properties (interoperability and self-sovereignty). We explored how these elements could align and then highlight research gaps and present a DIdM research agenda.

Task structure and knowledge transfer: leveraging employee agility performance in an ESM environment

ABSTRACT The current study employs communication visibility theory to examine the nexus between task structure, knowledge transfer, and employee agility. We further investigated the role of ESM in facilitating knowledge transfer and its effects on employee agility. Utilising cross-sectional data from Chinese employees working in different organisations, structural equation modelling (SEM) was applied to 425 samples using AMOS version 24.0. Results reveal that task interdependence significantly influences both knowledge-sharing and knowledge-hiding behaviours in organisations. Moreover, task complexity positively affects knowledge sharing but negatively impacts knowledge hiding. The results also indicate that knowledge sharing positively correlates with employee agility, while knowledge hiding has a negative association with employee agility. Additionally, knowledge sharing mediates the relationship between task interdependence, task complexity, and employee agility, as does knowledge hiding. Furthermore, ESM usage strengthens the relationship between knowledge sharing and employee agility, although it does not moderate the link between knowledge hiding and employee agility. The study concludes with substantial theoretical and practical implications and directions for future research.

Syntactic Adaptation and Word Learning in 3‐ to 4‐Year‐Olds

AbstractIn a recent study, preschoolers adapted their syntactic expectations about a familiar phrase in French; this adaptation affected later word learning. In two experiments, we probed the generality of this finding by replicating the experiment and extending it to a different expression in English. We examined the ambiguous phrase the baby, which can be followed by nouns (the baby monkeys) or verbs (the baby sleeps). In induction trials, the baby consistently preceded either familiar nouns (noun condition) or verbs (verb condition). In later novel‐word trials, children in the verb condition were more likely to interpret novel words following the baby (The baby gorps!) as verbs than were children in the noun condition. In Experiment 2, a modified design isolated the effect of experience with the critical phrase from possible effects of task structure, and an added baseline condition showed the adaptation effect to be asymmetrical, suggesting frequency or surprisal effects on adaptation.

Neural activity ramps in frontal cortex signal extended motivation during learning

Learning requires the ability to link actions to outcomes. How motivation facilitates learning is not well understood. We designed a behavioral task in which mice self-initiate trials to learn cue-reward contingencies and found that the anterior cingulate region of the prefrontal cortex (ACC) contains motivation-related signals to maximize rewards. In particular, we found that ACC neural activity was consistently tied to trial initiations where mice seek to leave unrewarded cues to reach reward-associated cues. Notably, this neural signal persisted over consecutive unrewarded cues until reward-associated cues were reached, and was required for learning. To determine how ACC inherits this motivational signal we performed projection-specific photometry recordings from several inputs to ACC during learning. In doing so, we identified a ramp in bulk neural activity in orbitofrontal cortex (OFC)-to-ACC projections as mice received unrewarded cues, which continued ramping across consecutive unrewarded cues, and finally peaked upon reaching a reward-associated cue, thus maintaining an extended motivational state. Cellular resolution imaging of OFC confirmed these neural correlates of motivation, and further delineated separate ensembles of neurons that sequentially tiled the ramp. Together, these results identify a mechanism by which OFC maps out task structure to convey an extended motivational state to ACC to facilitate goal-directed learning.

Neural activity ramps in frontal cortex signal extended motivation during learning.

Learning requires the ability to link actions to outcomes. How motivation facilitates learning is not well understood. We designed a behavioral task in which mice self-initiate trials to learn cue-reward contingencies and found that the anterior cingulate region of the prefrontal cortex (ACC) contains motivation-related signals to maximize rewards. In particular, we found that ACC neural activity was consistently tied to trial initiations where mice seek to leave unrewarded cues to reach reward-associated cues. Notably, this neural signal persisted over consecutive unrewarded cues until reward-associated cues were reached, and was required for learning. To determine how ACC inherits this motivational signal we performed projection-specific photometry recordings from several inputs to ACC during learning. In doing so, we identified a ramp in bulk neural activity in orbitofrontal cortex (OFC)-to-ACC projections as mice received unrewarded cues, which continued ramping across consecutive unrewarded cues, and finally peaked upon reaching a reward-associated cue, thus maintaining an extended motivational state. Cellular resolution imaging of OFC confirmed these neural correlates of motivation, and further delineated separate ensembles of neurons that sequentially tiled the ramp. Together, these results identify a mechanism by which OFC maps out task structure to convey an extended motivational state to ACC to facilitate goal-directed learning.

Inductive biases of neural network modularity in spatial navigation.

The brain may have evolved a modular architecture for daily tasks, with circuits featuring functionally specialized modules that match the task structure. We hypothesize that this architecture enables better learning and generalization than architectures with less specialized modules. To test this, we trained reinforcement learning agents with various neural architectures on a naturalistic navigation task. We found that the modular agent, with an architecture that segregates computations of state representation, value, and action into specialized modules, achieved better learning and generalization. Its learned state representation combines prediction and observation, weighted by their relative uncertainty, akin to recursive Bayesian estimation. This agent's behavior also resembles macaques' behavior more closely. Our results shed light on the possible rationale for the brain's modularity and suggest that artificial systems can use this insight from neuroscience to improve learning and generalization in natural tasks.

A-star algorithm vs. Q-learning algorithm for path planning: A comparative analysis

In robotics research, the core focus is on path planning, as it's essential for robots to navigate according to human objectives to be valuable. This paper provides an in-depth analysis of the commonly used A-Star algorithm and Q-learning algorithm in mobile robot path planning. It specifically examines the advantages and disadvantages of both the A-Star algorithm and the Q-learning algorithm, highlighting their roles in the advancement of robotic intelligence. The A-star algorithm is an algorithm with a heuristic function to guide path selection direction. Also, the A-star algorithm is guaranteed to identify the most efficient path in terms of traffic cost when facing static obstacles. However, its performance degrades when dealing with trap nodes. Specifically, the algorithm occupies a large amount of memory and takes a long computational time. In contrast, the Q-learning algorithm is effective for learning continuous actions with the same network data structure for local path planning tasks.

Visual working memory models of delayed estimation do not generalize to whole-report tasks.

Whole-report working memory tasks provide a measure of recall for all stimuli in a trial and afford single-trial analyses that are not possible with single-report delayed estimation tasks. However, most whole-report studies assume that trial stimuli are encoded and reported independently, and they do not consider the relationships between stimuli presented and reported within the same trial. Here, we present the results of two independently conducted whole-report experiments. The first dataset was recorded by Adam, Vogel, and Awh (2017) and required participants to report color and orientation stimuli using a continuous response wheel. We recorded the second dataset, which required participants to report color stimuli using a set of discrete buttons. We found that participants often group their reports by color similarity, contradicting the assumption of independence implicit in most encoding models of working memory. Next, we showed that this behavior was consistent across participants and experiments when reporting color but not orientation, two circular variables often assumed to be equivalent.Finally, we implemented an alternative to independent encoding where stimuli are encoded as a hierarchical Bayesian ensemble and found that this model predicts biases that are not present in either dataset. Our results suggest that assumptions made by both independent and hierarchical ensemble encoding models-which were developed in the context of single-report delayed estimation tasks-do not hold for the whole-report task. This failure to generalize highlights the need to consider variations in task structure when inferring fundamental principles of visual working memory.

Temporally extended successor feature neural episodic control

One of the long-term goals of reinforcement learning is to build intelligent agents capable of rapidly learning and flexibly transferring skills, similar to humans and animals. In this paper, we introduce an episodic control framework based on the temporal expansion of subsequent features to achieve these goals, which we refer to as Temporally Extended Successor Feature Neural Episodic Control (TESFNEC). This method has shown impressive results in significantly improving sample efficiency and elegantly reusing previously learned strategies. Crucially, this model enhances agent training by incorporating episodic memory, significantly reducing the number of iterations required to learn the optimal policy. Furthermore, we adopt the temporal expansion of successor features a technique to capture the expected state transition dynamics of actions. This form of temporal abstraction does not entail learning a top-down hierarchy of task structures but focuses on the bottom-up combination of actions and action repetitions. Thus, our approach directly considers the temporal scope of sequences of temporally extended actions without requiring predefined or domain-specific options. Experimental results in the two-dimensional object collection environment demonstrate that the method proposed in this paper optimizes learning policies faster than baseline reinforcement learning approaches, leading to higher average returns.

Task structure tailors the geometry of neural representations in human lateral prefrontal cortex.

How do human brains represent tasks of varying structure? The lateral prefrontal cortex (lPFC) flexibly represents task information. However, principles that shape lPFC representational geometry remain unsettled. We use fMRI and pattern analyses to reveal the structure of lPFC representational geometries as humans perform two distinct categorization tasks- one with flat, conjunctive categories and another with hierarchical, context-dependent categories. We show that lPFC encodes task-relevant information with task-tailored geometries of intermediate dimensionality. These geometries preferentially enhance the separability of task-relevant variables while encoding a subset in abstract form. Specifically, in the flat task, a global axis encodes response-relevant categories abstractly, while category-specific local geometries are high-dimensional. In the hierarchy task, a global axis abstractly encodes the higher-level context, while low-dimensional, context-specific local geometries compress irrelevant information and abstractly encode the relevant information. Comparing these task geometries exposes generalizable principles by which lPFC tailors representations to different tasks.

Regulation of reinforcement learning parameters captures long-term changes in rat behaviour.

In uncertain environments in which resources fluctuate continuously, animals must permanently decide whether to stabilise learning and exploit what they currently believe to be their best option, or instead explore potential alternatives and learn fast from new observations. While such a trade-off has been extensively studied in pretrained animals facing non-stationary decision-making tasks, it is yet unknown how they progressively tune it while learning the task structure during pretraining. Here, we compared the ability of different computational models to account for long-term changes in the behaviour of 24 rats while they learned to choose a rewarded lever in a three-armed bandit task across 24 days of pretraining. We found that the day-by-day evolution of rat performance and win-shift tendency revealed a progressive stabilisation of the way they regulated reinforcement learning parameters. We successfully captured these behavioural adaptations using a meta-learning model in which either the learning rate or the inverse temperature was controlled by the average reward rate.

Фактори ефективності кооперативного навчання: теоретичні підходи і емпіричні дослідження

The article deals with the problem of factors and psychological mechanisms of influence of cooperative learning on students' achievement. On the basis of comparative analysis, four approaches to understanding the factors of effectiveness of cooperative learning are distinguished: motivational, socio-psychological, cognitive-genetic, and cognitive-functional. The motivational approach (D. Johnson, R. Johnson, R. Slavin) brings to the fore the motivation of performing joint educational tasks, determining the structure of rewards or group educational tasks that stimulate and direct the activity of students. The socio-psychological approach (E. Aronson, V. Battisch, D. Solomon, K. Delucci, E. Cohen, S. Sharan, Y. Sharan) emphasizes the cohesion of students, their interpersonal relationships, which depend on the success of educational activities: students actively participate in group tasks and help each other if they identify with the group and strive for joint success. Proponents of this approach attach great importance to ways of forming student groups (teams), establishing constructive interpersonal relationships during joint learning, as well as reflecting on the process and results of joint cognitive activity. Proponents of cognitive approaches believe that interaction between students in itself increases their success due to intensive cognitive processing of information, rather than motivation or social cohesion of students. Two slightly different cognitive approaches are distinguished, which developed in parallel ways: cognitive-genetic and cognitive-functional. The cognitive-genetic approach, which was formed in line with the theories of mental development of L. Vygotsky and J. Piaget, is based on the assumption that communicative interaction in the process of performing cognitive tasks contributes to the mental development of students and increases the quality of assimilation of educational material (M. Burns, B. Wadsworth, W. Damon, F. Murray). The cognitive-functional approach arose in the vein of cognitive psychology and claims that the positive effect of cooperative learning on the success of students is ensured by cognitive restructuring (processing) of educational information, which contributes to its preservation in memory and integration into the previously learned system of knowledge (S. Allen, N. Webb, M. Wittrock, D. Dansereau, L. Devin-Sheehan, A. O'Donnell, A. Palincsar, M. Presley, R. Feldman). Summarizing the results of research in the field of cooperative learning makes it possible to create a coherent theoretical model that integrates various approaches and comprehensively considers the main factors of the effectiveness of cooperative learning: motivation, group cohesion, processes of interpersonal interaction that stimulate cognitive development and processing of educational material. In such a model, each of the four approaches considered plays a role and makes an important contribution to understanding the interrelated processes of cooperative learning.

Resource Redeployment and the Pursuit of the New Best Use: Economic Logic and Organizational Challenges

Shifting economic circumstances may suggest some degree of resource redeployment of capacity-constrained capabilities. However, a separate question remains as to how corporate actors operate within the space of this economic logic. While prior literature on diversification has tended to highlight the fungibility of resources as a basis for their redeployment, we point to an additional consideration of the complementarity of resources that makes allocation within the firm attractive, what we term a “new best use.” Further, we point to a potential tension as the economic logic suggested by this property of complementarity, and more generally interdependence with the firm’s other resources and activities, may make assessing what constitutes what we term the “new best use” of a resource challenging. The latent fungibility of a resource may differ from the range of reliable assessment of relative value creation. The allocation of resources and capabilities across a firm of any scale or scope is generally guided by the organizational and task structure of the enterprise. Building on these arguments, we consider two classes of factors that might impact the realization of the latent opportunities of resource redeployment. One is the “pipes” of organizational structures and grouping of business units and allocation of decision rights of specific managers to those units, and the other is the “prisms” of the various metrics of performance and value creation that may make resource use across different domains of the business more or less comparable.

FairGAT: Fairness-Aware Graph Attention Networks

Graphs can facilitate modeling various complex systems such as gene networks and power grids as well as analyzing the underlying relations within them. Learning over graphs has recently attracted increasing attention, particularly graph neural network (GNN)–based solutions, among which graph attention networks (GATs) have become one of the most widely utilized neural network structures for graph-based tasks. Although it is shown that the use of graph structures in learning results in the amplification of algorithmic bias, the influence of the attention design in GATs on algorithmic bias has not been investigated. Motivated by this, the present study first carries out a theoretical analysis in order to demonstrate the sources of algorithmic bias in GAT-based learning for node classification. Then, a novel algorithm, FairGAT, which leverages a fairness-aware attention design, is developed based on the theoretical findings. Experimental results on real-world networks demonstrate that FairGAT improves group fairness measures while also providing comparable utility to the fairness-aware baselines for node classification and link prediction.