The Experts below are selected from a list of 1365 Experts worldwide ranked by ideXlab platform

Silvia A Bunge - One of the best experts on this subject based on the ideXlab platform.

  • hemispheric differences in Relational Reasoning novel insights based on an old technique
    Frontiers in Human Neuroscience, 2015
    Co-Authors: Michael S Vendetti, Elizabeth L Johnson, Connor J Lemos, Silvia A Bunge
    Abstract:

    Relational Reasoning, or the ability to integrate multiple mental relations to arrive at a logical conclusion, is a critical component of higher cognition. A bilateral brain network involving lateral prefrontal and parietal cortices has been consistently implicated in Relational Reasoning. Some data suggest a preferential role for the left hemisphere in this form of Reasoning, whereas others suggest that the two hemispheres make important contributions. To test for a hemispheric asymmetry in Relational Reasoning, we made use of an old technique known as visual half-field stimulus presentation to manipulate whether stimuli were presented briefly to one hemisphere or the other. Across two experiments, 54 neurologically healthy young adults performed a visuospatial transitive inference task. Pairs of colored shapes were presented rapidly in either the left or right visual hemifield as participants maintained central fixation, thereby isolating initial encoding to the contralateral hemisphere. We observed a left-hemisphere advantage for encoding a series of ordered visuospatial relations, but both hemispheres contributed equally to task performance when the relations were presented out of order. To our knowledge, this is the first study to reveal hemispheric differences in Relational encoding in the intact brain. We discuss these findings in the context of a rich literature on hemispheric asymmetries in cognition.

  • Is Relational Reasoning dependent on language? A voxel-based lesion symptom mapping study
    Brain and Language, 2010
    Co-Authors: Juliana V. Baldo, Stephen M Wilson, Silvia A Bunge, N. F. Dronkers
    Abstract:

    Previous studies with brain-injured patients have suggested that language abilities are necessary for complex problem-solving, even when tasks are non-verbal. In the current study, we tested this notion by analyzing behavioral and neuroimaging data from a large group of left-hemisphere stroke patients (n=107) suffering from a range of language impairment from none to severe. Patients were tested on the Raven's Colored Progressive Matrices (RCPM), a non-verbal test of Reasoning that requires participants to complete a visual pattern or sequence with one of six possible choices. For some items, the solution could be determined by visual pattern-matching, but other items required more complex, Relational Reasoning. As predicted, performance on the Relational-Reasoning items was disproportionately affected in language-impaired patients with aphasia, relative to non-aphasic, left-hemisphere patients. A voxel-based lesion symptom mapping (VLSM) procedure was used to relate patients' RCPM performance with areas of damage in the brain. Results showed that deficits on the Relational Reasoning problems were associated with lesions in the left middle and superior temporal gyri, regions essential for language processing, as well as in the left inferior parietal lobule. In contrast, the visual pattern-matching condition was associated with lesions in posterior portions of the left hemisphere that subserve visual processing, namely, occipital and inferotemporal cortex. These findings provide compelling support for the idea that language is critical for higher-level Reasoning and problem-solving. © 2010.

  • neurocognitive development of Relational Reasoning
    Developmental Science, 2009
    Co-Authors: Silvia A Bunge, Eveline A Crone, Carter Wendelken, Linda Van Leijenhorst, Ryan D Honomichl, Kalina Christoff
    Abstract:

    Relational Reasoning is an essential component of fluid intelligence, and is known to have a protracted developmental trajectory. To date, little is known about the neural changes that underlie improvements in Reasoning ability over development. In this event-related functional magnetic resonance imaging (fMRI) study, children aged 8‐12 and adults aged 18‐25 performed a Relational Reasoning task adapted from Raven’s Progressive Matrices. The task included three levels of Relational Reasoning demands: REL-0, REL-1, and REL-2. Children exhibited disproportionately lower accuracy than adults on trials that required integration of two relations (REL-2). Like adults, children engaged lateral prefrontal cortex (PFC) and parietal cortex during task performance; however, they exhibited different time courses and activation profiles, providing insight into their approach to the problems. As in prior studies, adults exhibited increased rostrolateral PFC (RLPFC) activation when Relational integration was required (REL-2 > REL-1, REL-0). Children also engaged RLPFC most strongly for REL-2 problems at early stages of processing, but this differential activation relative to REL-1 trials was not sustained throughout the trial. These results suggest that the children recruited RLPFC while processing relations, but failed to use it to integrate across two relations. Relational integration is critical for solving a variety of problems, and for appreciating analogies; the current findings suggest that developmental improvements in this function rely on changes in the profile of engagement of RLPFC, as well as dorsolateral PFC and parietal cortex.

  • brain is to thought as stomach is to investigating the role of rostrolateral prefrontal cortex in Relational Reasoning
    Journal of Cognitive Neuroscience, 2008
    Co-Authors: Carter Wendelken, Denis Nakhabenko, Sarah E. Donohue, Cameron S. Carter, Silvia A Bunge
    Abstract:

    Brain imaging studies suggest that the rostrolateral prefrontal cortex (RLPFC), is involved in Relational Reasoning. Functional magnetic resonance imaging (fMRI) studies involving Raven's Progressive Matrices or verbal propositional analogies indicate that the RLPFC is engaged by tasks that require integration across multiple Relational structures. Several studies have shown that the RLPFC is more active when people must evaluate an analogy (e.g., Is shoe to foot as glove is to hand?) than when they must simply evaluate two individual semantic relationships, consistent with the hypothesis that this region is important for Relational integration. The current fMRI investigation further explores the role of the RLPFC in Reasoning and Relational integration by comparing RLPFC activation across four different propositional analogy conditions. Each of the four conditions required either relation completion (e.g., Shoe is to foot as glove is to WHAT? hand) or relation comparison (e.g., Is shoe to foot as glove is to hand? yes). The RLPFC was engaged more strongly by the comparison subtask relative to completion, suggesting that the RLPFC is particularly involved in comparing Relational structures.

  • “Brain Is to Thought as Stomach Is to ??”: Investigating the Role of Rostrolateral Prefrontal Cortex in Relational Reasoning
    Journal of Cognitive Neuroscience, 2008
    Co-Authors: Carter Wendelken, Denis Nakhabenko, Sarah E. Donohue, Cameron S. Carter, Silvia A Bunge
    Abstract:

    Brain imaging studies suggest that the rostrolateral prefrontal cortex (RLPFC), is involved in Relational Reasoning. Functional magnetic resonance imaging (fMRI) studies involving Raven's Progressive Matrices or verbal propositional analogies indicate that the RLPFC is engaged by tasks that require integration across multiple Relational structures. Several studies have shown that the RLPFC is more active when people must evaluate an analogy (e.g., Is shoe to foot as glove is to hand?) than when they must simply evaluate two individual semantic relationships, consistent with the hypothesis that this region is important for Relational integration. The current fMRI investigation further explores the role of the RLPFC in Reasoning and Relational integration by comparing RLPFC activation across four different propositional analogy conditions. Each of the four conditions required either relation completion (e.g., Shoe is to foot as glove is to WHAT? → “hand”) or relation comparison (e.g., Is shoe to foot as glove is to hand? → “yes”). The RLPFC was engaged more strongly by the comparison subtask relative to completion, suggesting that the RLPFC is particularly involved in comparing Relational structures.

Patricia A. Alexander - One of the best experts on this subject based on the ideXlab platform.

  • individual differences in college age learners the importance of Relational Reasoning for learning and assessment in higher education
    British Journal of Educational Psychology, 2019
    Co-Authors: Patricia A. Alexander
    Abstract:

    BACKGROUND: The term individual differences refers to the physical, behavioral, cognitive, social, and emotional attributes that make each human unique. Late adolescence to young adulthood represents a time of significant neurobiological and cognitive transformations that contribute further to human variability. Those transformations include an increase in the white matter of the brain accompanied by an increased capacity for higher-order thinking, Reasoning, decision-making, and selfcontrol. These capacities fall under the category of executive functions. AIM: The purpose of this article is to overview one particular executive function, Relational Reasoning; to consider its significance to college students' learning and performance; and to argue for its inclusion in assessment programs within higher education. RESEARCH SUMMARY: Relational Reasoning can be defined as the ability to discern meaningful patterns within any informational stream. Through an orchestrated program of research, four forms of Relational Reasoning have been identified that reflect perceived similarities (analogical), discrepancies (anomalous), contradictions (antinomous), and contrasts (antithetical). These four forms have been documented in studies of doctors diagnosing cases, science and mathematics teachers providing instruction, and engineering students designing new products. These manifestations have also formed the structure of formal measures of Relational Reasoning that have been shown to be psychometrically sound indicators of this executive function. CONCLUSIONS AND IMPLICATIONS: This article closes with an argument for the inclusion of Relational Reasoning in performance assessments designed for higher education. Consistent with views of competence as a continuum, it is argued that this inclusion should encompass the measurement of Relational Reasoning both as an underlying general competence and as a component of domain-specific performance.

  • The development of Relational Reasoning in primary and secondary school students: a longitudinal investigation in technology education
    International Journal of Technology and Design Education, 2019
    Co-Authors: Sophie Jablansky, Patricia A. Alexander, Denis Dumas, Vicki Compton
    Abstract:

    For several decades, there has been a push to advance students’ knowledge and abilities in science, technology, engineering, and mathematics (STEM). One capacity that has been linked positively to STEM achievement is Relational Reasoning, which involves identifying associations between objects, ideas, and situations. Yet, few studies have examined Relational Reasoning and its component forms (i.e., analogy, anomaly, antinomy, antithesis) within the domain of technology or how these abilities might change over time. The present study explored the development of primary and secondary school students’ Relational Reasoning over a period of 2 years as they interacted with technological objects. Participants (n = 59) were a subset of a nationally representative random sample between 5 and 18 years old. Students met with a researcher to discuss the form and function of a familiar and unfamiliar technological object at two time points. Results demonstrated that students of all ages used Relational Reasoning to identify associations between objects’ functionality and form, but that the types and amounts of Relational Reasoning varied by grade group, time, and object familiarity. This study has implications for researchers and practitioners interested in the development of Relational Reasoning and technological literacy, and suggests possible ways of enhancing both.

  • Individual Differences in College-Age Learners: The Importance of Relational Reasoning for Learning and Assessment in Higher Education.
    British Journal of Educational Psychology, 2019
    Co-Authors: Patricia A. Alexander
    Abstract:

    The term individual differences refers to the physical, behavioral, cognitive, social, and emotional attributes that make each human unique. Late adolescence to young adulthood represents a time of significant neurobiological and cognitive transformations that contribute further to human variability. Those transformations include an increase in the white matter of the brain accompanied by an increased capacity for higher-order thinking, Reasoning, decision-making, and selfcontrol. These capacities fall under the category of executive functions. The purpose of this article is to overview one particular executive function, Relational Reasoning; to consider its significance to college students' learning and performance; and to argue for its inclusion in assessment programs within higher education. Relational Reasoning can be defined as the ability to discern meaningful patterns within any informational stream. Through an orchestrated program of research, four forms of Relational Reasoning have been identified that reflect perceived similarities (analogical), discrepancies (anomalous), contradictions (antinomous), and contrasts (antithetical). These four forms have been documented in studies of doctors diagnosing cases, science and mathematics teachers providing instruction, and engineering students designing new products. These manifestations have also formed the structure of formal measures of Relational Reasoning that have been shown to be psychometrically sound indicators of this executive function. This article closes with an argument for the inclusion of Relational Reasoning in performance assessments designed for higher education. Consistent with views of competence as a continuum, it is argued that this inclusion should encompass the measurement of Relational Reasoning both as an underlying general competence and as a component of domain-specific performance. © 2019 The British Psychological Society.

  • assessing differential item functioning on the test of Relational Reasoning
    Frontiers in Education, 2018
    Co-Authors: Denis Dumas, Patricia A. Alexander
    Abstract:

    The Test of Relational Reasoning (TORR) is designed to assess the ability to identify complex patterns within visuospatial stimuli. The TORR is designed for use in school and university settings, and therefore, its measurement invariance across diverse groups is critical. In this investigation, a large sample, representative of a major U.S. university on key demographic variables, was collected, and the resulting data were analyzed using a multi-group, multidimensional item-response theory model-comparison procedure. No significant differential item functioning (DIF) was found on any of the TORR items across any of the demographic groups of interest. This finding is interpreted as evidence of the cultural fairness of the TORR, and potential test-development choices that may have contributed to that cultural fairness are discussed.

  • Relational Reasoning in STEM Domains: a Foundation for Academic Development
    Educational Psychology Review, 2017
    Co-Authors: Patricia A. Alexander
    Abstract:

    What is Relational Reasoning? Why is it critical to consider the role of Relational Reasoning in students learning and development in science, technology, engineering, and mathematics (STEM)? Moreover, how do the particular contributions populating this special issue address the pressing societal needs and offer guidance to researchers and practitioners concerned with students academic development in these essential domains? These are the questions that are explored in this introduction to this special issue, Relational Reasoning in STEM Domains: What Empirical Research Can Contribute to the National Dialogue.

Iroise Dumontheil - One of the best experts on this subject based on the ideXlab platform.

  • FAST-TRACK ARTICLE Development of Relational Reasoning during adolescence
    2020
    Co-Authors: Iroise Dumontheil, Kalina Christoff, Rachael Houlton, Sarahjayne Blakemore
    Abstract:

    Non-linear changes in behaviour and in brain activity during adolescent development have been reported in a variety of cognitive tasks. These developmental changes are often interpreted as being a consequence of changes in brain structure, including nonlinear changes in grey matter volumes, which occur during adolescence. However, very few studies have attempted to combine behavioural, functional and structural data. This multi-method approach is the one we took in the current study, which was designed to investigate developmental changes in behaviour and brain activity during Relational Reasoning, the simultaneous integration of multiple relations. We used a Relational Reasoning task known to recruit rostrolateral prefrontal cortex (RLPFC), a region that undergoes substantial structural changes during adolescence. The task was administered to female participants in a behavioural (N = 178, 7–27 years) and an fMRI study (N = 37, 11–30 years). Non-linear changes in accuracy were observed, with poorer performance during mid-adolescence. fMRI and VBM results revealed a complex picture of linear and possibly nonlinear changes with age. Performance and structural changes partly accounted for changes with age in RLPFC and medial superior frontal gyrus activity but not for a decrease in activation in the anterior insula ⁄ frontal operculum between midadolescence and adulthood. These functional changes might instead reflect the maturation of neurocognitive strategies.

  • Social and Nonsocial Relational Reasoning in Adolescence and Adulthood
    Journal of Cognitive Neuroscience, 2017
    Co-Authors: Lucía Magis-weinberg, Sarahjayne Blakemore, Iroise Dumontheil
    Abstract:

    Reasoning during social interactions requires the individual manipulation of mental representations of one's own traits and those of other people as well as their joint consideration (Relational integration). Research using nonsocial paradigms has linked Relational integration to activity in the rostrolateral PFC. Here, we investigated whether social Reasoning is supported by the same general system or whether it additionally relies on regions of the social brain network, such as the medial PFC. We further assessed the development of social Reasoning. In the social task, participants evaluated themselves or a friend, or compared themselves with their friend, on a series of traits. In the nonsocial task, participants evaluated their hometown or another town or compared the two. In a behavioral study involving 325 participants (11–39 years old), we found that integrating relations, compared with performing single Relational judgments, improves during adolescence, both for social and nonsocial information. Thirty-nine female participants (10–31 years old) took part in a neuroimaging study using a similar task. Activation of the Relational integration network, including the rostrolateral PFC, was observed in the comparison condition of both the social and nonsocial tasks, whereas the medial PFC showed greater activation when participants processed social as opposed to nonsocial information across conditions. Developmentally, the right anterior insula showed greater activity in adolescents compared with adults during the comparison of nonsocial versus social information. This study shows parallel recruitment of the social brain and the Relational Reasoning network during the Relational integration of social information in adolescence and adulthood.

  • Audience effects on the neural correlates of Relational Reasoning in adolescence
    Neuropsychologia, 2016
    Co-Authors: Iroise Dumontheil, Laura K. Wolf, Sarahjayne Blakemore
    Abstract:

    Adolescents are particularly sensitive to peer influence. This may partly be due to an increased salience of peers during adolescence. We investigated the effect of being observed by a peer on a cognitively challenging task, Relational Reasoning, which requires the evaluation and integration of multiple mental representations. Relational Reasoning tasks engage a fronto-parietal network including the inferior parietal cortex, pre-supplementary motor area, dorsolateral and rostrolateral prefrontal cortices. Using functional magnetic resonance imaging (fMRI), peer audience effects on activation in this fronto-parietal network were compared in a group of 19 female mid-adolescents (aged 14–16 years) and 14 female adults (aged 23–28 years). Adolescent and adult Relational Reasoning accuracy was influenced by a peer audience as a function of task difficulty: the presence of a peer audience led to decreased accuracy in the complex, Relational integration condition in both groups of participants. The fMRI results demonstrated that a peer audience differentially modulated activation in regions of the fronto-parietal network in adolescents and adults. Activation was increased in adolescents in the presence of a peer audience, while this was not the case in adults.

  • developmental changes in effective connectivity associated with Relational Reasoning
    Human Brain Mapping, 2014
    Co-Authors: Narges Bazargani, Kalina Christoff, Iroise Dumontheil, Hauke Hillebrandt
    Abstract:

    Rostrolateral prefrontal cortex (RLPFC) is part of a frontoparietal network of regions involved in Relational Reasoning, the mental process of working with relationships between multiple mental repre- sentations. RLPFC has shown functional and structural changes with age, with increasing specificity of left RLPFC activation for Relational integration during development. Here, we used dynamic causal mod- eling (DCM) to investigate changes in effective connectivity during a Relational Reasoning task through the transition from adolescence into adulthood. We examined fMRI data of 37 healthy female participants (11-30 years old) performing a Relational Reasoning paradigm. Comparing Relational integration to the manipulation of single relations revealed activation in five regions: the RLPFC, anterior insula, dorsolat- eral PFC, inferior parietal lobe, and medial superior frontal gyrus. We used a new exhaustive search approach and identified a full DCM model, which included all reciprocal connections between the five clusters in the left hemisphere, as the optimal model. In line with previous resting state fMRI results, we showed distinct developmental effects on the strength of long-range frontoparietal versus frontoinsular short-range fixed connections. The modulatory connections associated with Relational integration increased with age. Gray matter volume in left RLPFC, which decreased with age, partly accounted for changes in fixed PFC connectivity. Finally, improvements in Relational integration performance were associated with greater modulatory and weaker fixed PFC connectivity. This pattern provides further evidence of increas- ing specificity of left PFC function for Relational integration compared to the manipulation of single rela- tions, and demonstrates an association between effective connectivity and performance during development. Hum Brain Mapp 35:3262-3276, 2014. V C 2013 Wiley Periodicals, Inc.

  • development of Relational Reasoning during adolescence
    Developmental Science, 2010
    Co-Authors: Kalina Christoff, Iroise Dumontheil, Rachael Houlton, Sarahjayne Blakemore
    Abstract:

    Non-linear changes in behaviour and in brain activity during adolescent development have been reported in a variety of cognitive tasks. These developmental changes are often interpreted as being a consequence of changes in brain structure, including non-linear changes in grey matter volumes, which occur during adolescence. However, very few studies have attempted to combine behavioural, functional and structural data. This multi-method approach is the one we took in the current study, which was designed to investigate developmental changes in behaviour and brain activity during Relational Reasoning, the simultaneous integration of multiple relations. We used a Relational Reasoning task known to recruit rostrolateral prefrontal cortex (RLPFC), a region that undergoes substantial structural changes during adolescence. The task was administered to female participants in a behavioural (N = 178, 7-27 years) and an fMRI study (N = 37, 11-30 years). Non-linear changes in accuracy were observed, with poorer performance during mid-adolescence. fMRI and VBM results revealed a complex picture of linear and possibly non-linear changes with age. Performance and structural changes partly accounted for changes with age in RLPFC and medial superior frontal gyrus activity but not for a decrease in activation in the anterior insula/frontal operculum between mid-adolescence and adulthood. These functional changes might instead reflect the maturation of neurocognitive strategies.

Marco Ragni - One of the best experts on this subject based on the ideXlab platform.

  • Conditional Relational Reasoning - eScholarship
    2020
    Co-Authors: Marco Ragni, Tobias Sonntag
    Abstract:

    Conditional Relational Reasoning Marco Ragni University of Freiburg Tobias Sonntag University of Freiburg Abstract: Conditional Reasoning - Reasoning with ”if” - is very common in everyday life. Consequently, all Reasoning theories provide explanations for typical logical errors in conditional Reasoning. The combination of conditional Reasoning with spatial Relational Reasoning provides a new test-bed for these theories. Take, for instance, the two assertions ”If the church is to the left of the cinema then the cinema is to the left of the library” and ”The church is to the right of the library”. Can both assertion be true at the same time? The second statement links a term from the antecedent to the consequence. Such novel tasks pose significant problems for the reasoner, and extensions of classical theories differ in their empirical predictions. We report two behavioral experiments investigating conditional Relational Reasoning and provide support for predictions of the mental model theory.

  • uncertain Relational Reasoning in the parietal cortex
    Brain and Cognition, 2016
    Co-Authors: Marco Ragni, Imke Franzmeier, Simon Maier, Markus Knauff
    Abstract:

    The psychology of Reasoning is currently transitioning from the study of deductive inferences under certainty to inferences that have degrees of uncertainty in both their premises and conclusions; however, only a few studies have explored the cortical basis of uncertain Reasoning. Using transcranial magnetic stimulation (TMS), we show that areas in the right superior parietal lobe (rSPL) are necessary for solving spatial Relational Reasoning problems under conditions of uncertainty. Twenty-four participants had to decide whether a single presented order of objects agreed with a given set of indeterminate premises that could be interpreted in more than one way. During the presentation of the order, 10-Hz TMS was applied over the rSPL or a sham control site. Right SPL TMS during the inference phase disrupted performance in uncertain Relational Reasoning. Moreover, we found differences in the error rates between preferred mental models, alternative models, and inconsistent models. Our results suggest that different mechanisms are involved when people reason spatially and evaluate different kinds of uncertain conclusions.

  • Premise annotation in mental model construction: An ACT-R approach to processing indeterminacy in spatial Relational Reasoning
    Cognitive Systems Research, 2013
    Co-Authors: Sven Brüssow, Marco Ragni, Matthias Frorath, Lars Konieczny, Thomas Fangmeier
    Abstract:

    Reasoning about inference problems that allow for multiple interpretations requires maintaining intermediate representations that, if necessary, may be reconsidered at a later stage of the solution process. In that respect we describe the process of annotating premises in spatial Relational Reasoning that allows for the derivation of alternative representations. Furthermore, we show how ACT-R's subsymbolic processing principles substantially contribute to the underlying theoretical framework of the Preferred Mental Model Theory as they add a powerful component making precise accuracy predictions possible, a feature that in previous symbolic approaches has been neglected. In addition, we implemented and compared two strategies to investigate the persistence of the outcomes of the Reasoning process. Furthermore, we examined how well data and predictions meet the central assumption that Reasoning difficulty increases with the number of mental operations necessary to validate a putative conclusion.

  • Spatial Cognition - Linguistic principles for spatial Relational Reasoning
    Spatial Cognition VIII, 2012
    Co-Authors: Thora Tenbrink, Marco Ragni
    Abstract:

    Human spatial Relational Reasoning has been investigated by presenting participants with premises like: "The triangle is to the left of the circle, the circle is to the left of the square. Which relation holds between the triangle and the square?" Participants are expected to interpret the descriptions in a way that corresponds to the logical options that are theoretically available. Recent findings on spatial language usage highlight a range of pragmatic principles that speakers intuitively adhere to when producing and comprehending spatial relationship descriptions; these appear to contradict the principles used in Relational Reasoning studies. In order to clarify the relation between speakers' intuitions and the descriptions used in Relational Reasoning tasks, we present two studies in which linguistic representations of relevant configurations were elicited. Results highlight the systematic patterns speakers use in describing these configurations, adding new insights to research on spatial language usage across various levels of analysis. We argue that the identified principles may interfere with the Reasoning processes investigated in earlier studies, and suggest that future studies should adequately account for the principles underlying intuitive spatial language use.

Timothy Lillicrap - One of the best experts on this subject based on the ideXlab platform.

  • NIPS - A simple neural network module for Relational Reasoning
    2020
    Co-Authors: Adam Santoro, David Raposo, David G T Barrett, Mateusz Malinowski, Razvan Pascanu, Peter W Battaglia, Timothy Lillicrap
    Abstract:

    Relational Reasoning is a central component of generally intelligent behavior, but has proven difficult for neural networks to learn. In this paper we describe how to use Relation Networks (RNs) as a simple plug-and-play module to solve problems that fundamentally hinge on Relational Reasoning. We tested RN-augmented networks on three tasks: visual question answering using a challenging dataset called CLEVR, on which we achieve state-of-the-art, super-human performance; text-based question answering using the bAbI suite of tasks; and complex Reasoning about dynamical physical systems. Then, using a curated dataset called Sort-of-CLEVR we show that powerful convolutional networks do not have a general capacity to solve Relational questions, but can gain this capacity when augmented with RNs. Thus, by simply augmenting convolutions, LSTMs, and MLPs with RNs, we can remove computational burden from network components that are not well-suited to handle Relational Reasoning, reduce overall network complexity, and gain a general ability to reason about the relations between entities and their properties.

  • a simple neural network module for Relational Reasoning
    arXiv: Computation and Language, 2017
    Co-Authors: Adam Santoro, David Raposo, David G T Barrett, Mateusz Malinowski, Razvan Pascanu, Peter W Battaglia, Timothy Lillicrap
    Abstract:

    Relational Reasoning is a central component of generally intelligent behavior, but has proven difficult for neural networks to learn. In this paper we describe how to use Relation Networks (RNs) as a simple plug-and-play module to solve problems that fundamentally hinge on Relational Reasoning. We tested RN-augmented networks on three tasks: visual question answering using a challenging dataset called CLEVR, on which we achieve state-of-the-art, super-human performance; text-based question answering using the bAbI suite of tasks; and complex Reasoning about dynamic physical systems. Then, using a curated dataset called Sort-of-CLEVR we show that powerful convolutional networks do not have a general capacity to solve Relational questions, but can gain this capacity when augmented with RNs. Our work shows how a deep learning architecture equipped with an RN module can implicitly discover and learn to reason about entities and their relations.

  • a simple neural network module for Relational Reasoning
    Neural Information Processing Systems, 2017
    Co-Authors: Adam Santoro, David Raposo, David G T Barrett, Mateusz Malinowski, Razvan Pascanu, Peter W Battaglia, Timothy Lillicrap
    Abstract:

    Relational Reasoning is a central component of generally intelligent behavior, but has proven difficult for neural networks to learn. In this paper we describe how to use Relation Networks (RNs) as a simple plug-and-play module to solve problems that fundamentally hinge on Relational Reasoning. We tested RN-augmented networks on three tasks: visual question answering using a challenging dataset called CLEVR, on which we achieve state-of-the-art, super-human performance; text-based question answering using the bAbI suite of tasks; and complex Reasoning about dynamical physical systems. Then, using a curated dataset called Sort-of-CLEVR we show that powerful convolutional networks do not have a general capacity to solve Relational questions, but can gain this capacity when augmented with RNs. Thus, by simply augmenting convolutions, LSTMs, and MLPs with RNs, we can remove computational burden from network components that are not well-suited to handle Relational Reasoning, reduce overall network complexity, and gain a general ability to reason about the relations between entities and their properties.