A mathematical theory of relational generalization in transitive inference.

Humans and animals routinely infer relations between different items or events and generalize these relations to novel combinations of items. This allows them to respond appropriately to radically novel circumstances and is fundamental to advanced cognition. However, how learning systems (including the brain) can implement the necessary inductive biases has been unclear. We investigated transitive inference (TI), a classic relational task paradigm in which subjects must learn a relation ([Formula: see text] and [Formula: see text]) and generalize it to new combinations of items ([Formula: see text]). Through mathematical analysis, we found that a broad range of biologically relevant learning models (e.g. gradient flow or ridge regression) perform TI successfully and recapitulate signature behavioral patterns long observed in living subjects. First, we found that models with item-wise additive representations automatically encode transitive relations. Second, for more general representations, a single scalar "conjunctivity factor" determines model behavior on TI and, further, the principle of norm minimization (a standard statistical inductive bias) enables models with fixed, partly conjunctive representations to generalize transitively. Finally, neural networks in the "rich regime," which enables representation learning and improves generalization on many tasks, unexpectedly show poor generalization and anomalous behavior on TI. We find that such networks implement a form of norm minimization (over hidden weights) that yields a local encoding mechanism lacking transitivity. Our findings show how minimal statistical learning principles give rise to a classical relational inductive bias (transitivity), explain empirically observed behaviors, and establish a formal approach to understanding the neural basis of relational abstraction.

Chained study and the discovery of relational structure.

Prior knowledge of relational structure allows people to quickly make sense of and respond to new experiences. When awareness of such structure is not necessary to support learning, however, it is unclear when and why individuals "spontaneously discover" an underlying relational schema. The present study examines the determinants of such discovery in discrimination-based transitive inference (TI), whereby people learn about a hierarchy of interrelated premises and are tested on their ability to draw inferences that bridge studied relations. Experiencing "chained" sequences of overlapping premises during training was predicted to facilitate the discovery of relational structure. Among individuals without prior knowledge of the hierarchy, chaining improved relational learning and was most likely to result in explicit awareness of the underlying relations between items. Observation of chained training sequences was also more effective than the self-generation of training sequences. These findings add to growing evidence that the temporal dynamics of training, including successive presentation of overlapping associations, are key to understanding spontaneous relational discovery during learning.

Use of temporal and colour cueing in a symbolic delayed matching task by honey bees.

Honey bees (Apis mellifera) are known for their capacity to learn arbitrary relationships between colours, odours and even numbers. However, it is not known whether bees can use temporal signals as cueing stimuli in a similar way during symbolic delayed matching-to-sample tasks. Honey bees potentially process temporal signals during foraging activities, but the extent to which they can use such information is unclear. Here, we investigated whether free-flying honey bees could use either illumination colour or illumination duration as potential context-setting cues to enable their subsequent decisions for a symbolic delayed matching-to-sample task. We found that bees could use the changing colour context of the illumination to complete the subsequent spatial vision task at a level significantly different from chance expectation, but could not use the duration of either a 1 or 3 s light as a cueing stimulus. These findings suggest that bees cannot use temporal information as a cueing stimulus as efficiently as other signals such as colour, and are consistent with previous field observations suggesting a limited interval timing capacity in honey bees.

A qualitative evaluation of the effect of a longitudinal dementia education programme on healthcare student knowledge and attitudes.

BACKGROUND AND OBJECTIVES: There is a need to ensure that the future healthcare workforce has the necessary knowledge and skills to deliver high quality compassionate care to the increasing number of people with dementia. Our programme has been set up to address this challenge. In the programme, undergraduate healthcare students (nursing, medical and paramedic) visit a family (person with dementia and their carer) in pairs over a 2-year period. This qualitative study sought to understand the student experience of the programme. METHODS: Participants were undergraduate healthcare students who were undertaking our programme at two universities. We sampled for variation in the student participants in order to generate a framework for understanding the student experience of the programme. Students were invited to take part in the qualitative study, and written consent was obtained. Interviews and focus group transcripts were analysed using thematic analysis. RESULTS: Thirty-nine (nursing, medical and paramedic) student participants took part in individual in-depth qualitative interviews and 38 took part in five focus groups. Four key themes were identified from the analysis; relational learning, insight and understanding, challenging attitudes and enhanced dementia practice. DISCUSSION: Student experience of our programme was shown to be positive. The relationship between the students and family was most impactful in supporting student learning, and the subsequent improvement in knowledge, attitudes and practice. Our model of undergraduate dementia education has applicability for other long-term conditions.

Relational Scaffolding Enhances Children's Understanding of Scientific Models.

Models are central to the practice and teaching of science. Yet people often fail to grasp how scientific models explain their observations of the world. Realizing the explanatory power of a model may require aligning its relational structure to that of the observable phenomena. In the present study, we tested whether relational scaffolding-guided comparisons between observable and modeled events-enhances children's understanding of scientific models. We tested relational scaffolding during instruction of third graders about the day/night cycle, a topic that involves relating Earth-based observations to a space-based model of Earth's rotation. Experiment 1 found that participants (N = 108) learned more from instruction that incorporated relational scaffolding. Experiment 2 (N = 99) found that guided comparison-not merely viewing observable and modeled events-is a critical component of relational scaffolding, especially for children with low initial knowledge. Relational scaffolding could be applied broadly to assist the many students who struggle with science.

A Decision-Support Tool for Renal Mass Classification.

We investigate the viability of statistical relational machine learning algorithms for the task of identifying malignancy of renal masses using radiomics-based imaging features. Features characterizing the texture, signal intensity, and other relevant metrics of the renal mass were extracted from multiphase contrast-enhanced computed tomography images. The recently developed formalism of relational functional gradient boosting (RFGB) was used to learn human-interpretable models for classification. Experimental results demonstrate that RFGB outperforms many standard machine learning approaches as well as the current diagnostic gold standard of visual qualification by radiologists.

Markov Logic Networks for Adverse Drug Event Extraction from Text.

Adverse drug events (ADEs) are a major concern and point of emphasis for the medical profession, government, and society. A diverse set of techniques from epidemiology, statistics, and computer science are being proposed and studied for ADE discovery from observational health data (e.g., EHR and claims data), social network data (e.g., Google and Twitter posts), and other information sources. Methodologies are needed for evaluating, quantitatively measuring, and comparing the ability of these various approaches to accurately discover ADEs. This work is motivated by the observation that text sources such as the Medline/Medinfo library provide a wealth of information on human health. Unfortunately, ADEs often result from unexpected interactions, and the connection between conditions and drugs is not explicit in these sources. Thus, in this work we address the question of whether we can quantitatively estimate relationships between drugs and conditions from the medical literature. This paper proposes and studies a state-of-the-art NLP-based extraction of ADEs from text.

Acquisition of colour categories through learning: Differences between hue and lightness.

Colour categories are acquired through learning, but the nature of this process is not fully understood. Some category distinctions are defined by hue (e.g. red/purple) but other by lightness (red/pink). The aim of this study was to investigate if the acquisition of key information for making accurate cross-boundary discriminations poses different challenges for hue-defined as opposed to lightness-defined boundaries. To answer this question, hue- and lightness-learners were trained on a novel category boundary within the GREEN region of colour space. After training, hue- and lightness-learners as well as untrained controls performed delayed same-different discrimination for lightness and hue pairs. In addition to discrimination data, errors during learning and category-labelling strategies were examined. Errors during learning distributed non-uniformly and in accordance with the Bezold-Brücke effect, which accounts for darker colours at the green-blue boundary appearing greener and lighter colours appearing bluer. Only hue-learners showed discrimination improvements due to category boundary acquisition. Thus, acquisition is more efficient for hue-category compared to lightness-category boundaries. Almost all learners reported using category-labelling strategies, with hue-learners almost exclusively using 'green'/'blue' and lightness learners using a wider range of labels, most often 'light'/'dark'. Thus, labels play an important role in colour category learning and such labelling does not conform to everyday naming: here, the label 'blue' is used for exemplars that would normally be named 'green'. In conclusion, labelling serves the purpose of highlighting key information that differentiates exemplars across the category boundary, and basic colour terms may be particularly effective in facilitating such attentional guidance.

Logic + probabilistic programming + causal laws.

Probabilistic planning attempts to incorporate stochastic models directly into the planning process, which is the problem of synthesizing a sequence of actions that achieves some objective for a putative agent. Probabilistic programming has rapidly emerged as a key paradigm to integrate probabilistic concepts with programming languages, which allows one to specify complex probabilistic models using programming primitives like recursion and loops. Probabilistic logic programming aims to further ease the specification of structured probability distributions using first-order logical artefacts. In this article, we briefly discuss the modelling of probabilistic planning through the lens of probabilistic (logic) programming. Although many flavours for such an integration are possible, we focus on two representative examples. The first is an extension to the popular probabilistic logic programming language PROBLOG, which permits the decoration of probabilities on Horn clauses-that is, prolog programs. The second is an extension to the popular agent programming language GOLOG, which permits the logical specification of dynamical systems via actions, effects and observations. The probabilistic extensions thereof emphasize different strengths of probabilistic programming that are particularly useful for non-trivial modelling issues raised in probabilistic planning. Among other things, one can instantiate planning problems with growing and shrinking state spaces, discrete and continuous probability distributions, and non-unique prior distributions in a first-order setting.

Relational reasoning network for anatomical landmarking.

Purpose: We perform anatomical landmarking for craniomaxillofacial (CMF) bones without explicitly segmenting them. Toward this, we propose a simple, yet efficient, deep network architecture, called relational reasoning network (RRN), to accurately learn the local and the global relations among the landmarks in CMF bones; specifically, mandible, maxilla, and nasal bones. Approach: The proposed RRN works in an end-to-end manner, utilizing learned relations of the landmarks based on dense-block units. For a given few landmarks as input, RRN treats the landmarking process similar to a data imputation problem where predicted landmarks are considered missing. Results: We applied RRN to cone-beam computed tomography scans obtained from 250 patients. With a fourfold cross-validation technique, we obtained an average root mean squared error of < 2 mm per landmark. Our proposed RRN has revealed unique relationships among the landmarks that help us in inferring informativeness of the landmark points. The proposed system identifies the missing landmark locations accurately even when severe pathology or deformations are present in the bones. Conclusions: Accurately identifying anatomical landmarks is a crucial step in deformation analysis and surgical planning for CMF surgeries. Achieving this goal without the need for explicit bone segmentation addresses a major limitation of segmentation-based approaches, where segmentation failure (as often is the case in bones with severe pathology or deformation) could easily lead to incorrect landmarking. To the best of our knowledge, this is the first-of-its-kind algorithm finding anatomical relations of the objects using deep learning.

Learning and reasoning with graph data.

Reasoning about graphs, and learning from graph data is a field of artificial intelligence that has recently received much attention in the machine learning areas of graph representation learning and graph neural networks. Graphs are also the underlying structures of interest in a wide range of more traditional fields ranging from logic-oriented knowledge representation and reasoning to graph kernels and statistical relational learning. In this review we outline a broad map and inventory of the field of learning and reasoning with graphs that spans the spectrum from reasoning in the form of logical deduction to learning node embeddings. To obtain a unified perspective on such a diverse landscape we introduce a simple and general semantic concept of a model that covers logic knowledge bases, graph neural networks, kernel support vector machines, and many other types of frameworks. Still at a high semantic level, we survey common strategies for model specification using probabilistic factorization and standard feature construction techniques. Based on this semantic foundation we introduce a taxonomy of reasoning tasks that casts problems ranging from transductive link prediction to asymptotic analysis of random graph models as queries of different complexities for a given model. Similarly, we express learning in different frameworks and settings in terms of a common statistical maximum likelihood principle. Overall, this review aims to provide a coherent conceptual framework that provides a basis for further theoretical analyses of respective strengths and limitations of different approaches to handling graph data, and that facilitates combination and integration of different modeling paradigms.

MRChexNet: Multi-modal bridge and relational learning for thoracic disease recognition in chest X-rays.

While diagnosing multiple lesion regions in chest X-ray (CXR) images, radiologists usually apply pathological relationships in medicine before making decisions. Therefore, a comprehensive analysis of labeling relationships in different data modes is essential to improve the recognition performance of the model. However, most automated CXR diagnostic methods that consider pathological relationships treat different data modalities as independent learning objects, ignoring the alignment of pathological relationships among different data modalities. In addition, some methods that use undirected graphs to model pathological relationships ignore the directed information, making it difficult to model all pathological relationships accurately. In this paper, we propose a novel multi-label CXR classification model called MRChexNet that consists of three modules: a representation learning module (RLM), a multi-modal bridge module (MBM) and a pathology graph learning module (PGL). RLM captures specific pathological features at the image level. MBM performs cross-modal alignment of pathology relationships in different data modalities. PGL models directed relationships between disease occurrences as directed graphs. Finally, the designed graph learning block in PGL performs the integrated learning of pathology relationships in different data modalities. We evaluated MRChexNet on two large-scale CXR datasets (ChestX-Ray14 and CheXpert) and achieved state-of-the-art performance. The mean area under the curve (AUC) scores for the 14 pathologies were 0.8503 (ChestX-Ray14) and 0.8649 (CheXpert). MRChexNet effectively aligns pathology relationships in different modalities and learns more detailed correlations between pathologies. It demonstrates high accuracy and generalization compared to competing approaches. MRChexNet can contribute to thoracic disease recognition in CXR.

Mapping principles and worked examples for structural learning: effects of content complexity.

Drawing connections between principles and worked examples is an approach to learning and instruction, but it is poorly understood. This study investigated the effects of principle and example complexity on learners' ability to map principles and worked examples. The complexity of a principle or example was determined based on the number of concepts and relationships involved. 138 college students were randomly assigned to one of the mapping conditions: principle-simple example, principle-complex example, simple example-simple example, simple example-complex example, and complex example-complex example. The participants studied related materials and completed a free-mapping and a guided-mapping task for a simple and a complex probability principle. The effects of the mapping activities were measured in terms of gains in structural and conceptual knowledge. For the simple principle, principle-example mapping led to fewer nonrelational comparisons (standalone concepts) than did example-example mapping and an equal number of relational comparisons (interconnected concepts). For the complex principle, principle-example mapping led to fewer nonrelational but more relational comparisons than example-example mapping did. Principle-example mapping of corresponding content was more difficult than example-example mapping was. However, principle-example mapping of noncorresponding content was as easy as or easier than example-example mapping. The two forms of mapping resulted in equivalent gains in structural and conceptual knowledge. The findings of this study expand the understanding of analogical reasoning and learning through mapping and comparison of abstract and concrete content. The findings indicate that principle-example mapping enables learners to overcome the obstacles of comprehending abstract or general information and to identify the interrelationships of the individual concepts in formal structures.

No evidence for language benefits in infant relational learning.

Recent studies have found that infants show relational learning in the first year. Like older children, they can abstract relations such as same or different across a series of exemplars. For older children, language has a major impact on relational learning: labeling a shared relation facilitates learning, while labeling component objects can disrupt learning. Here we ask: Does language influence relational learning at 12 months? Experiment 1 (n = 64) examined the influence of a relational label on learning. Prior to the study, the infants saw three pairs of objects, all labeled "These are same" or "These are different". Experiment 2 (n = 48) examined the influence of object labels prior to the study, with three objects labeled (e.g., "This is a cup, this is a tower."). We compared the present results with those of Ferry et al. (2015), where infants abstracted same and different relations after undergoing a similar paradigm without prior labels. If the effects of language mirror those in older children, we would expect that infants given relational labels (Experiment 1) will be helped in abstracting same and different compared to infants not given labels and that infants given object labels (Experiment 2) will be hindered relative to those not given labels. We found no evidence for either prediction. In Experiment 1, infants who had heard relational labels did not benefit compared to infants who had received no labels (Ferry et al., 2015). In Experiment 2, infants who had heard object labels showed the same patterns as those in Ferry et al. (2015), suggesting that object labels had no effect. This finding is important because it highlights a key difference between the relational learning abilities of infants and those seen in older children, pointing to a protracted process by which language and relational learning become entwined.

The effects of testing the relationships among relational concepts.

Many concepts are defined by their relationships to one another. However, instructors might teach these concepts individually, neglecting their interconnections. For instance, students learning about statistical power might learn how to define alpha and beta, but not how they are related. We report two experiments that examine whether there is a benefit to training subjects on relations among concepts. In Experiment 1, all subjects studied material on statistical hypothesis testing, half were subsequently quizzed on relationships among these concepts, and the other half were quizzed on their individual definitions; quizzing was used to highlight the information that was being trained in each condition (i.e., relations or definitions). Experiment 2 also included a mixed training condition that quizzed both relations and definitions, and a control condition that only included study. Subjects were then tested on both types of questions and on three conceptually related question types. In Experiment 1, subjects trained on relations performed numerically better on relational test questions than subjects trained on definitions (nonsignificant trend), whereas definitional test questions showed the reverse pattern; no performance differences were found between the groups on the other question types. In Experiment 2, relational training benefitted performance on relational test questions and on some question types that were not quizzed, whereas definitional training only benefited performance on test questions on the trained definitions. In contrast, mixed training did not aid learning above and beyond studying. Relational training thus seems to facilitate transfer of learning, whereas definitional training seems to produce training specificity effects.

Co-design and delivery of a relational learning programme for nursing students and young people with severe and complex learning disabilities.

BACKGROUND: National Health Service (NHS) staff in the UK are required to undergo training about learning disabilities at the appropriate level for their role. However, this requirement does not apply to nurses in training and student nurses report fear and anxiety about caring for people with intellectual learning disabilities (ILDs). Young people with intellectual disabilities report feeling scared of nurses and parents feel staff do not listen to them or involve them in care. OBJECTIVES: (i) For a university and special school for young people with server and complex ILDs to work in partnership to co-design a programme for nursing students, young people, their teachers, and parents. (ii) To deliver the programme online as part of the university's existing nursing courses. DESIGN: The partnership between the university and the special school focused on co-design of an interactive programme, parent involvement, safeguarding, and the design of accessible learning resources to support young people with severe and complex ILDs' engagement. The programme was informed by relational inquiry, service user and transformative pedagogies, and parents and teacher's knowledge and views about the young people. Delivery of the programme was designed to fit into existing nursing courses and enable students on placement and young people at home or in hospital to participate. A rights-based ethnographic evaluation was designed to support participant feedback and programme development. SETTING: The Heritage2Health Virtual Arts and Drama Programme was piloted with nursing students at one UK university and young people with severe and complex ILDs from one special school, their parents and teachers. PARTICIPANTS: 15 nursing students (BSc Year 2 = 10, Year 3 = 3, MSc = 2) and 7 young people with severe and complex learning disabilities (age 11-14 yrs). Other participants were parents/guardians of young people (7), arts/drama facilitators (2), academic lecturers (2), special needs teachers (2), registered nurse (1). METHODS: An 8-week dynamic programme of arts and drama. Sessions included 30-min start-up/presencing, 45-min storytelling/drama with young people and parents, 30-min reflection/close. Sessions were facilitated by 2 arts and drama specialists. The story of 'Ubuntu the Lion with the Long, Long, Mane' (by TNP) was used to explore difference and ways of being. The evaluation methods were participant observation, semi-structured interviews (2-6 weeks post) and thematic analysis. RESULTS: Participation in the programme was a challenging, creative, and reflective experience that was transformative for all. Nurses and young people's fears and anxieties about each other were revealed and addressed by participating in arts and drama activities together. Nursing students learnt how to adopt a relational orientation to young people and their parents and teachers. CONCLUSIONS: A co-designed programme for nursing students and young people with severe and complex ILDs can benefit student knowledge and skills and reduce fears and anxieties between nurses and young people with ILDs. With adequate planning and resources, the programme could be adopted by multidisciplinary partnerships between other universities and special schools.

Designing Exhibits to Support Relational Learning in a Science Museum.

Science museums aim to provide educational experiences for both children and adults. To achieve this goal, museum displays must convey scientifically-relevant relationships, such as the similarities that unite members of a natural category, and the connections between scientific models and observable objects and events. In this paper, we explore how research on comparison could be leveraged to support learning about such relationships. We describe how museum displays could promote educationally-relevant comparisons involving natural specimens and scientific models. We also discuss how these comparisons could be supported through the design of a display-in particular, by using similarity, space, and language to facilitate relational thinking for children and their adult companions. Such supports may be pivotal given the informal nature of learning in museums.

Relational Learning Improves Prediction of Mortality in COVID-19 in the Intensive Care Unit.

Traditional Machine Learning (ML) models have had limited success in predicting Coronoavirus-19 (COVID-19) outcomes using Electronic Health Record (EHR) data partially due to not effectively capturing the inter-connectivity patterns between various data modalities. In this work, we propose a novel framework that utilizes relational learning based on a heterogeneous graph model (HGM) for predicting mortality at different time windows in COVID-19 patients within the intensive care unit (ICU). We utilize the EHRs of one of the largest and most diverse patient populations across five hospitals in major health system in New York City. In our model, we use an LSTM for processing time varying patient data and apply our proposed relational learning strategy in the final output layer along with other static features. Here, we replace the traditional softmax layer with a Skip-Gram relational learning strategy to compare the similarity between a patient and outcome embedding representation. We demonstrate that the construction of a HGM can robustly learn the patterns classifying patient representations of outcomes through leveraging patterns within the embeddings of similar patients. Our experimental results show that our relational learning-based HGM model achieves higher area under the receiver operating characteristic curve (auROC) than both comparator models in all prediction time windows, with dramatic improvements to recall.

Propositionalization and embeddings: two sides of the same coin.

Data preprocessing is an important component of machine learning pipelines, which requires ample time and resources. An integral part of preprocessing is data transformation into the format required by a given learning algorithm. This paper outlines some of the modern data processing techniques used in relational learning that enable data fusion from different input data types and formats into a single table data representation, focusing on the propositionalization and embedding data transformation approaches. While both approaches aim at transforming data into tabular data format, they use different terminology and task definitions, are perceived to address different goals, and are used in different contexts. This paper contributes a unifying framework that allows for improved understanding of these two data transformation techniques by presenting their unified definitions, and by explaining the similarities and differences between the two approaches as variants of a unified complex data transformation task. In addition to the unifying framework, the novelty of this paper is a unifying methodology combining propositionalization and embeddings, which benefits from the advantages of both in solving complex data transformation and learning tasks. We present two efficient implementations of the unifying methodology: an instance-based PropDRM approach, and a feature-based PropStar approach to data transformation and learning, together with their empirical evaluation on several relational problems. The results show that the new algorithms can outperform existing relational learners and can solve much larger problems.

Elemental and Configural Associative Learning in Spatial Tasks: Could Zebrafish be Used to Advance Our Knowledge?

Spatial learning and memory have been studied for several decades. Analyses of these processes pose fundamental scientific questions but are also relevant from a biomedical perspective. The cellular, synaptic and molecular mechanisms underlying spatial learning have been intensively investigated, yet the behavioral mechanisms/strategies in a spatial task still pose unanswered questions. Spatial learning relies upon configural information about cues in the environment. However, each of these cues can also independently form part of an elemental association with the specific spatial position, and thus spatial tasks may be solved using elemental (single CS and US association) learning. Here, we first briefly review what we know about configural learning from studies with rodents. Subsequently, we discuss the pros and cons of employing a relatively novel laboratory organism, the zebrafish in such studies, providing some examples of methods with which both elemental and configural learning may be explored with this species. Last, we speculate about future research directions focusing on how zebrafish may advance our knowledge. We argue that zebrafish strikes a reasonable compromise between system complexity and practical simplicity and that adding this species to the studies with laboratory rodents will allow us to gain a better understanding of both the evolution of and the mechanisms underlying spatial learning. We conclude that zebrafish research will enhance the translational relevance of our findings.