30472 research outputs found
Sort by
Selective particle attention: rapidly and flexibly selecting features for deep reinforcement learning
Get PDFDeep Reinforcement Learning (RL) is often criticized for being data inefficient and in
exible to changes
in task structure. Part of the reason for these issues is that Deep RL typically learns end-to-end using
backpropagation, which results in task-specifc representations. One approach for circumventing these
problems is to apply Deep RL to existing representations that have been learned in a more task-agnostic
fashion. However, this only partially solves the problem as the Deep RL algorithm learns a function of all
pre-existing representations and is therefore still susceptible to data inefficiency and a lack of
exibility.
Biological agents appear to solve this problem by forming internal representations over many tasks and
only selecting a subset of these features for decision-making based on the task at hand; a process commonly
referred to as selective attention. We take inspiration from selective attention in biological agents and propose
a novel algorithm called Selective Particle Attention (SPA), which selects subsets of existing representations
for Deep RL. Crucially, these subsets are not learned through backpropagation, which is slow and prone to
overfitting, but instead via a particle filter that rapidly and
exibly identifies key subsets of features using
only reward feedback. We evaluate SPA on two tasks that involve raw pixel input and dynamic changes to
the task structure, and show that it greatly increases the efficiency and
exibility of downstream Deep RL
algorithms
Introduction: the future of global Byzantium
No full textBook synopsis: Global Byzantium is, in part, a recasting and expansion of the old ‘Byzantium and its neighbours’ theme with, however, a methodological twist away from the resolutely political and toward the cultural and economic. A second thing that Global Byzantium – as a concept – explicitly endorses is comparative methodology. Global Byzantium needs also to address three further issues: cultural capital, the importance of the local, and the Empire’s strategic geographical location. Cultural capital: in past decades it was fashionable to define Byzantium as culturally superior to western Christian Europe, and Byzantine influence was a key concept, especially in art historical circles. This concept has been increasingly criticised, and what we now see emerging is a comparative methodology that relies on the concept of ‘competitive sharing’, not blind copying but rather competitive appropriation. The importance of the local is equally critical. We need to talk more about what the Byzantines saw when they ‘looked out’, and what others saw in Byzantium when they ‘looked in’ and to think about how that impacted on our, very post-modern, concepts of globalism. Finally, we need to think about the Empire’s strategic geographical position: between the 4th and the 13th centuries, if anyone was travelling internationally they had to travel across (or along the coasts of) the Byzantine Empire. Byzantium was thus a crucial intermediary, for good or for ill, between Europe, Africa and Asia – effectively, the glue that held the Christian world together, and it was also a critical transit point between the various Islamic polities and the Christian world
IR's sea sickness: a materialist diagnosis
Get PDFThis paper adopts a materialist understanding of nature, suggesting the ‘malaise’ of the discipline when it comes to the maritime factor in International Relations lies in the lack of attention to the terraqueous predicament of our planet. It elaborates on the implications of such an approach with reference to temporal and spatial aspects of modern international relations. Understanding the global ocean as a dynamic, living force in constant and changing interaction with terrestrial power can deliver deeper and richer interpretations of international relations, including conflicts like those in the contemporary South China Sea.
Book synopsis: While the world's oceans cover more than seventy percent of its surface, the sea has largely vanished as an object of enquiry in International Relations (IR), being treated either as a corollary of land or as time. Yet, the sea is the quintessential international space, and its importance to global politics has become all the more obvious in recent years. Drawing on interdisciplinary insights from IR, Historical Sociology, Blue Humanities and Critical Ocean Studies, The sea and International Relations breaks with this trend of oceanic amnesia, and kickstarts a theoretical, conceptual and empirical discussion about the sea and IR, by highlighting theoretical puzzles, analysing broad historical perspectives and addressing contemporary challenges. In bringing the sea back into IR, the book reconceptualises the canvas of international relations to include the oceans as a social, political, economic and military space which affects the workings of world politics
Did you know that you can deposit the version of record of Bloomsbury Academic book chapters in your institutional repository?
No full textWide-angle image rectification: a survey
Get PDFWide field-of-view (FOV) cameras, which capture a larger scene area than narrow FOV cameras, are used in many applications including 3D reconstruction, autonomous driving, and video surveillance. However, wide-angle images contain distortions that violate the assumptions underlying pinhole camera models, resulting in object distortion, difficulties in estimating scene distance, area, and direction, and preventing the use of off-the-shelf deep models trained on undistorted images for downstream computer vision tasks. Image rectification, which aims to correct these distortions, can solve these problems. In this paper we comprehensively survey progress in wide-angle image rectification from transformation models to rectification methods. Specifically, we first present a detailed description and discussion of the camera models used in different approaches. Then we summarize several distortion models including radial distortion and projection distortion. Next, we review both traditional geometry-based image rectification methods and deep learning based methods, where the former formulates disortion parameter estimation as an optimization problem and the latter treats it as a regression problem by leveraging the power of deep neural networks. We evaluate the performance of state-of-the-art methods on public databases and show that although both kinds of methods can achieve good results, these methods only work well for specific camera models and distortion types. We also provide a strong baseline model and carry out an empirical study of different distortion models on synthetic datasets and real-world wide-angle images. Finally, we discuss several potential research directons that are expected to further advance this area in the future
The macroeconomic cost of climate volatility
Get PDFWe study the impact of climate volatility on economic growth exploiting data on 133 countries between 1960 and 2019. We show that the conditional (exante) volatility of annual temperatures increased steadily over time, rendering climate conditions less predictable across countries, with important implications for growth. Controlling for concomitant changes in temperatures, a +1oC increase in temperature volatility causes on average a 0.3 per cent decline in GDP growth and a 0.7 per cent increase in the volatility of GDP. Unlike changes in average temperatures, changes in temperature volatility affect both rich and poor countries
Extreme weather, climate variability and childhood: A historical analogue from the Orkney Islands (1903-1919)
Get PDFMany small island communities are said to possess high levels of autonomous
coping capacity, often linked to peripherality. This social resilience is dynamic rather than
static, with environmental, social, and political drivers shaping local pattens of
vulnerability, necessitating reflection on how choices in one area may potentially lead to
new vulnerabilities or transfers of vulnerability to already sensitive groups, such as
children. This article argues that a historical perspective can help shed light on these
dynamics. Impacts of extreme weather and climate variability, and resultant impacts of
community coping strategies, on children in early-20th-century Orkney are explored using
school logbooks. It finds that extreme weather ‘shocks’ directly impact children’s ability
to attend school, while adjustments to the school calendar for agricultural operations
constitute an indirect impact of climate variability, with reduced recreation time an
emergent effect. Contextualised amidst contemporary island scholarship, two key
messages emerge. Firstly, that the mobility and/or work of children in island communities
remain sensitive to climate stressors in the present day and, secondly, that the island
context itself matters, as characteristics commonly associated with ‘islandness’ — such as
smallness, remoteness, and high social capital — may intersect in ways that fundamentally
impact children’s experiences of weather, work, and education
Transformation archetypes in global food systems
Get PDFFood systems are primary drivers of human and environmental health, but the understanding of their dynamic co-transformation remains limited. We use a data-driven approach to disentangle different development pathways of national food systems (i.e., ‘transformation archetypes’) based on historical, intertwined trends of food system structure (agricultural inputs and outputs and food trade), and social and environmental outcomes (malnutrition, biosphere integrity, and greenhouse gases emissions) for 161 countries, from 1995 to 2015. We found that whilst food systems have consistently improved in terms of productivity (ratio of output to input), other metrics suggest a typology of three transformation archetypes across countries: rapidly expansionist, expansionist, and consolidative. Expansionist and rapidly expansionist archetypes increased in agricultural area, synthetic fertiliser use, and gross agricultural output, which was accompanied by malnutrition, environmental pressures, and lasting socioeconomic disadvantages. The lowest rates of change in key structure metrics were found in the consolidative archetype. Across all transformation archetypes, agricultural greenhouse gases emissions, synthetic fertiliser use, and ecological footprint of consumption increased faster than the expansion of agricultural area, and obesity levels increased more rapidly than undernourishment decreased. The persistence of these unsustainable trajectories occurred independently of improvements in productivity. Our model underscores the importance of quantifying the multiple human and environmental dimensions of food systems transformations and can serve as a starting point to identify potential leverage points for sustainability transformations. More attention is thus warranted to alternative development pathways able of delivering equitable benefits to both productivity and to human and environmental health
Data sources:
Access Repository Dashboard
Do you manage Birkbeck Institutional Research Online? Access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard!