179566 research outputs found
Sort by
On the hidden layer-to-layer topology of the representations of reality realised within neural networks
Get PDFPurpose: Consider an information processing algorithm that is designed to process an input data object onto an output data object via a number of successive internal {\it layers} and mappings between them. The possible activation state within each layer can be represented as a cube within Euclidean space of a high dimension (e.g. equal to the number of artificial neurons at that level). Multiple instances of such input objects produce a point cloud within each layer’s cube: this is the “representation of the reality” at that layer, as sampled by the set of input objects. Design/methodology/approach: Most neural networks reduce the dimension of each layer’s cube from layer to successive layer. This gives the false impression of refining the inner representations of reality, distilling it down to fewer dimensions from which to discriminate or to infer outcomes (whatever is the aim). However, the representation of reality realised within each layer’s cube is a manifold, a curved subset embedded within it and of much lower dimension. Investigations show that such manifolds may not always be reducing in their local dimension. Instead, the manifold may become folded over and over, filling up further dimensions and creating non-realistic (unforeseeable) proximities. Findings: We discuss some of the likely consequences of these relatively unforeseen characteristics and, in particular, the possible vulnerability of such algorithms to non-realistic perturbations. We consider a possible response to this issue. Practical implications: New forms of calibration are necessary, using geometric/topological loss functions, as opposed to simple (variation-limiting) regularisation terms. Originality/value: We apply persistent homology methods to understand how the images of the point cloud (representing the sampled reality) change as they pass from layer to layer.</p
Low oxygen but dynamic marine redox conditions permitted the Cambrian Radiation
Get PDFWhether metazoan diversification during the Cambrian Radiation was driven by increased marine oxygenation remains highly debated. Repeated global oceanic oxygenation events have been inferred during this interval, but the degree of shallow marine oxygenation and its relationship to biodiversification and clade appearance remain uncertain. To resolve this, we interrogate an interval from ~527 to 519 Ma, encompassing multiple proposed global oceanic oxygenation events. We integrate the spatial and temporal distribution of shallow water, in situ reef metazoans, and trilobites, with high-resolution multi-proxy redox data through the highly biodiverse Siberian Platform. We document primarily dysoxic water column conditions, suggesting that early Cambrian metazoans, including motile skeletal benthos, had low oxygen demands. We further document oxygenation events coincident with positive carbon isotope excursions that led to modestly elevated oxygen levels. These events correspond to regional increases in species richness and habitat expansion of mainly endemic species, offering a potentially globally applicable model for biodiversification during the Cambrian Radiation
Modeling Disease Dynamics From Spatially Explicit Capture-Recapture Data
Get PDFOne of the main aims of wildlife disease ecology is to identify how disease dynamics vary in space and time and as a function of population density. However, monitoring spatiotemporal and density-dependent disease dynamics in the wild is challenging because the observation process is error-prone, which means that individuals, their disease status, and their spatial locations are unobservable, or only imperfectly observed. In this paper, we develop a novel spatially-explicit capture-recapture (SCR) model motivated by an SCR data set on European badgers (Meles meles), naturally infected with bovine tuberculosis (Mycobacterium bovis, TB). Our model accounts for the observation process of individuals as a function of their latent activity centers, and for their imperfectly observed disease status and its effect on demographic rates and behavior. This framework has the advantage of simultaneously modeling population demographics and disease dynamics within a spatial context. It can therefore generate estimates of critical parameters such as population size; local and global density by disease status and hence spatially-explicit disease prevalence; disease transmission probabilities as functions of local or global population density; and demographic rates as functions of disease status. Our findings suggest that infected badgers have lower survival probability but larger home range areas than uninfected badgers, and that the data do not provide strong evidence that density has a non-zero effect on disease transmission. We also present a simulation study, considering different scenarios of disease transmission within the population, and our findings highlight the importance of accounting for spatial variation in disease transmission and individual disease status when these affect demographic rates. Collectively these results show our new model enables a better understanding of how wildlife disease dynamics are linked to population demographics within a spatiotemporal context.</p
Participatory democracy and participatory research
Get PDFWhile normative theories of participatory democracy and practical experiences of participatory research share a common democratic commitment, the two fields have emerged and to date exist in isolation from each other. This article bridges this divide and asks what participatory democracy and participatory research can learn from one another. It argues that participatory democracy can learn how to realize its own democratic ideals within its research practice and participatory research can deepen its normative commitment by connecting its practices to a larger participatory vision. The article illustrates this by engaging with three examples in which participatory democracy researchers conduct participatory research projects. It finally reflects critically on how the shared participatory commitments of both fields can be realized within the neoliberal university embedded in competitive market economies.</p
Sorbent-based dialysate regeneration for the wearable artificial kidney: Advancing material innovation via experimental and computational studies
Get PDFHemodialysis is the primary renal replacement therapy for patients affected by end-stage renal disease, but it has a severe impact on the patient’s lifestyle and wellbeing, and is extremely water intensive. A wearable dialysis device could solve most of the issues associated with the treatment, but the main obstacle to its realisation is an efficient and reliable system for dialysate regeneration, i.e. the purification of spent dialysate from uremic toxins. Several techniques have been proposed to this aim, such as enzymatic conversion, forward osmosis and electrochemical oxidation. One of the most promising and safe technologies is adsorption, in which toxins are captured onto nanoporous materials, polymers or their combinations (mixed matrix membranes). In this review, we first give a general overview of the hemodialysis processes and of the challenges associated to making it wearable. Subsequently, we use experimental data from the literature to rank different materials based on their ability to remove the typical uremic toxins present in dialysate, including considerations on their recyclability, stability and safety. Finally, we critically analyse different computational modelling techniques available to design and/or optimise adsorbent materials for dialysate regeneration, and their accuracy in predicting the materials performance and screen large databases of adsorbents
Unravelling molecular mechanisms in atherosclerosis using cellular models and omics technologies
No full textDespite the discovery and prevalent clinical use of potent lipid-lowering therapies, including statins and PCSK9 inhibitors, cardiovascular diseases (CVD) caused by atherosclerosis remain a large unmet clinical need, accounting for frequent deaths worldwide. The pathogenesis of atherosclerosis is a complex process underlying the presence of modifiable and non-modifiable risk factors affecting several cell types including endothelial cells (ECs), monocytes/macrophages, smooth muscle cells (SMCs) and T cells. Heterogeneous composition of the plaque and its morphology could lead to rupture or erosion causing thrombosis, even a sudden death. To decipher this complexity, various cell model systems have been developed. With recent advances in systems biology approaches and single or multi-omics methods researchers can elucidate specific cell types, molecules and signalling pathways contributing to certain stages of disease progression. Compared with animals, in vitro models are economical, easily adjusted for high-throughput work, offering mechanistic insights. Hereby, we review the latest work performed employing the cellular models of atherosclerosis to generate a variety of omics data. We summarize their outputs and the impact they had in the field. Challenges in the translatability of the omics data obtained from the cell models will be discussed along with future perspectives
Cognitive rationality is heritable and lies under general cognitive ability
No full textIntelligence and rationality both predict optimal decision making. However, whether cognitive rationality (CR) and general cognitive ability (CA) are identical or reflect fundamentally distinct processes is hotly debated. Here, we report a twin study aimed at distinguishing the cognitive mechanisms involved in CR and CA. CR and CA tests were administered to a large twin sample. Univariate analyses indicated that both CA and CR were strongly heritable. Multivariate modelling of CA scales and CR indicated that CR was accounted for by a latent g-factor, which itself was strongly heritable. We conclude that CR is not a distinct disposition from CA, but instead that the reflexive and reflective aspects of cognitive ability make making CR a robust and efficient test of general cognitive ability
Recovery work in cascading and compounding disasters:A qualitative study of community recovery workers in Australia
No full textDisaster recovery has traditionally been conceptualised and operationalised on the basis of disasters occurring as single ‘events’ that are managed separately. However, there are cases where communities experience disasters that overlap, repeat, and compound in effects. This study addresses the following research questions: What do recovery workers see as the impacts of multiple disasters on community recovery? How is the process of supporting disaster recovery affected by another disaster, or disasters, occurring? Semi-structured qualitative interviews were conducted with recovery workers based in Australian communities that experienced two or more disasters between 2017 and 2022 (including drought, bushfires, landslides, and repeat floods). Recovery workers encountered interconnected impacts of disasters in the communities where they worked, including effects that accumulated, resurfaced and exacerbated each other. However, these workers experienced constraints in their roles, which were structured and funded in relation to a single specified disaster, hazard type, or ‘stage’ of disaster. This paper argues that events-based emergency management and its embedded disaster cycle of prevention, preparedness, response and recovery (PPRR) is at odds with the lived realities of recovery work after multiple disasters. Through making the case for moving past the framing of organisations and roles around single disasters, this paper underscores the need for new approaches to conceptualising, funding and structuring disaster recovery work for compounding and cascading disasters
Maternal education and prenatal smoking associations with adolescent executive function are substantially confounded by genetics
No full textTwin studies have suggested extremely high estimates of heritability for adolescent executive function, with no substantial contributions from shared environment. However, developmental psychology research has found significant correlations between executive function outcomes and elements of the environment that would be shared in twins. It is unclear whether these seemingly contradictory findings are best explained by genetic confounding in developmental studies, or limitations in twin studies which can potentially underestimate shared environment. In this study, we use genetic and phenotypic data from 5939 participants, 4827 participant mothers and 2903 participant fathers in the Millennium Cohort to examine the role of genetics in explaining common environmental associations with executive function, assessed by the spatial working memory task (SWM) and Cambridge Gambling task (CGT). Bivariate GCTA revealed that single nucleotide polymorphism (SNP) effects were the sole significant predictor of the association between SWM and both maternal education and prenatal smoking. M-GCTA and trioGCTA also found no significant evidence of indirect genetic effects on SWM, indicating that genetic nurture is unlikely to explain the bivariate GCTA results. The CGT showed no significant SNP heritability, suggesting that genetic influences on hot executive function may differ significantly from those on cool executive function. This study supports the twin study claim that the working memory component of executive function is primarily a genetic trait with minimal influence from shared environment, emphasizing the importance of using genetically sensitive designs to ensure that genetic confounding does not falsely inflate estimates of environmental influences on traits
Access Repository Dashboard
Do you manage Edinburgh Research Explorer? Access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard!