42337 research outputs found
Sort by
Three essays on asset pricing: from traditional models to neural networks
No full textIn the ever-evolving landscape of financial markets, understanding and predicting asset prices remain fundamental challenges in financial economics. This thesis advances the field of asset pricing by addressing two critical gaps. First, it investigates the momentum phenomenon in the options market, shedding light on the under-researched area and providing new empirical insights. Second, it bridges traditional asset pricing frameworks with advanced neural network models, exploring both cross-sectional and time-series contexts to capture complex, nonlinear, and time-varying relationships often overlooked by conventional methods.
The thesis comprises three distinct yet interconnected studies, each focusing on a specific aspect of asset pricing. These essays are unified by a shared progression, starting with traditional factor models and evolving toward the integration of conventional asset pricing frameworks with neural networks.
Chapter 2 examines the momentum phenomenon in the options market, focusing on three popular delta-neutral trading strategies: straddles, strangles, and butterfly spreads, using S&P 100 options data from 2000 to 2019. The study evaluates the robustness of momentum returns by accounting for traditional options risk factors, including the volatility spread, idiosyncratic volatility, market capitalization, the term spread of implied volatility, and the smirk slope of implied volatility. The findings reveal significant momentum returns
in both straddle and strangle strategies, which persist after controlling for these traditional risk factors and are unexplained by existing factor models. In contrast, no substantial momentum is observed for butterfly spreads, highlighting the strategy-specific nature of momentum effects in the options market.
Chapter 3 addresses the time series predictability of foreign exchange (FX) returns using neural networks. A Deep Prediction Model (DPRED) is introduced, incorporating cross-country interactions and long-memory dependencies. Empirical results indicate the presence of significant cross-country and long-memory effects, with the DPRED model providing improved forecasting accuracy. The findings also suggest that simpler neural network
architectures often yield better performance.
Chapter 4 examines neural network-based factor models for asset pricing in equity and foreign exchange (FX) markets, and develops a deep latent factor model that incorporates dynamic factor portfolios, extended historical asset dynamics, and cross-asset interactions. A series of progressively simplified deep learning models are designed to investigate the role of these components. The empirical study finds that FX models benefit from historical dynamics, while equity models perform better with dynamic factor weights. Furthermore, cross-asset interactions are significant in FX but minimal in equities. These findings highlight the critical role of tailored neural architectures in optimizing asset pricing models across different financial markets.
By integrating traditional asset pricing theories with advanced neural network methodologies, this thesis forms a cohesive narrative and contributes to both the theoretical and empirical asset pricing literature. It enhances our understanding of asset pricing dynamics across diverse financial markets, demonstrating the value of combining theoretical insights with machine learning techniques
Investigating chromosome ageing in mouse dictyate oocytes
Get PDFAgeing mammalian oocytes are known to be prone to transmit an abnormal set of chromosomes to the next generation, a condition called aneuploidy. One of the potential contributory factors is the loss of chromosome cohesion during the oocytes' prolonged meiotic arrest. Mature oocytes from aged mice have decreased amounts of cohesin, a structural protein that helps maintaining chromosome cohesion, associated with their chromosomes. However, these studies on the age-dependent loss of chromosome cohesion in oocytes were based on mature oocytes due to the challenging task of accessing pre-ovulatory oocytes. Therefore, it is currently unknown at which stage of oogenesis age-dependent loss of chromosome cohesion occurs. It is essential to understand when this age-dependent loss of oocyte cohesion happens to help us avoid, prevent or even treat the subsequent complications. Thus, the main objective of my thesis is to investigate when age-dependent chromosome cohesion loss occurs during oogenesis in mice.
Before investigating mouse oocytes, I tested a fluorescence in-situ hybridisation (FISH) method to monitor sister chromatid separation in a simpler experimental model to ensure the method’s compatibility with the detection of chromosome cohesion loss. FISH was carried out on Tex19.1 mouse embryonic stem cells (mES cells) that were previously shown to have defects in chromosome cohesion at the mitotic stage. Here I show that FISH signals hybridised to specific regions on mES cell chromosomes were significantly further apart in Tex19.1-/- mES cells, indicating increased sister chromatid separation. More importantly, FISH on mES has proved to be a promising method to study chromosome cohesion loss in mammalian cells.
To carry out a more in-depth study of chromosome cohesion in oocytes in fixed tissue, FISH was performed on Tex19.1 and Ubr2 mouse ovarian sections, two genes implicated in regulating cohesion in oocytes. However, there was no detectable increase in chromatid distances at the peri-telomere in Tex19.1-/- and Ubr2-/- oocytes in primordial follicles. It is possible that Tex19.1 and Ubr2 regulate cohesion at later stage of oogenesis. Notably, these experiments showed the capability of FISH to probe chromatid separation in meiotic cells in fixed tissue.
To understand the effect of ageing on chromatid cohesion in mouse oocytes at different follicular stages, I then performed FISH on 3-month, 6-month, and 9-month mouse ovarian tissue sections. Dictyate oocytes, in primordial, primary, and secondary follicular stages, were analysed for chromatid separation at the peri-centromeric and peri-telomeric regions of two chromosomes. These ageing oocytes showed significant increases in inter-chromatid distances at chromosome 1 from 3-month and 6-month to 9-month oocytes in all three follicular stages and at both peri-centromere and peri-telomere. Surprisingly, inter-chromatid distance increases were also detected during follicle development from primordial to secondary. However, no significant inter-chromatid distance change was detected in chromosome 18 of oocytes in primordial follicles. Taken together, although ageing may have different effects on chromosomes, age-dependent effects on chromosome arm separation were detected in dictyate oocytes.
In conclusion, these findings have provided new insights into our understanding of chromatid cohesion in female mouse germ cells and suggest that combining FISH with high-resolution 3D microscopy can allow the investigation of age-dependent changes in chromosome cohesion in dictyate oocytes. Ultimately, this research serves as a beneficial resource for the future study of ageing chromosome cohesion in pre-ovulatory oocytes in mammals, including humans
Classification of braces of order p⁴
Get PDFThe Yang-Baxter equation holds a fundamental place in both physics and mathematics, connecting to various fields such as knot theory, quantum integrable systems, quantum algebras, braid theory, and Hopf algebras, to name a few. In 2007 W. Rump introduced the brace, a generalisation of a nilpotent ring, as an algebraic tool for the study of solutions of the set-theoretic Yang-Baxter equation. More specifically, W. Rump showed that there is a one-to-one correspondence between braces and non-degenerate, involutive, set-theoretic solutions to the Yang-Baxter equation, thereby offering a pathway for finding novel solutions through the classification of braces. Since then, braces have been studied in connection to other algebraic structures, such as Hopf-Galois extensions, trusses, braided groups - and, of specific relevance to this thesis, pre-Lie rings and pre-Lie algebras. Pre-Lie rings and pre-Lie algebras are types of algebras over a ring whose product is left-symmetric. The connection between braces and pre-Lie algebras was first described in 2014 by W. Rump, in the case of R-braces and pre-Lie algebras over R. Later, in 2022, A. Smoktunowicz showed that right nilpotent braces of prime power cardinality correspond to pre-Lie rings of the same cardinality, for sufficiently large primes. In this case, right nilpotent braces can be explicitly obtained from the corresponding pre-Lie rings through the construction of the, so-called, group of flows. Hence, the classification of right nilpotent braces of prime power order can be achieved through the description of the corresponding pre-Lie rings. This thesis achieves the classification of braces of cardinality p⁴ for sufficiently large primes p⁴. The classification is separated into two parts, based on the right nilpotency of the braces under consideration. For braces of cardinality p⁴ that are right nilpotent, their correspondence to pre-Lie rings is leveraged. Working with pre-Lie rings is advantageous due to their comparative simplicity, primarily stemming from pre-Lie rings’ distributivity, which does not hold in braces. It is not known whether braces that are not right nilpotent correspond to pre-Lie rings, in which case we describe the braces directly, through a consideration of their defining generators and relations. We first determine the right nilpotency of braces of cardinality p⁴ for primes p such that p≥5. Then we describe all braces of cardinality p⁴ which are not right nilpotent for primes p such that p≥5. We also show that the constructed braces are prime, and contain a non-zero strongly nilpotent ideal. Using the constructed braces we produce examples of finite-dimensional pre-Lie algebras which are left nilpotent but not right nilpotent. Finally, we classify nilpotent pre-Lie rings of cardinality p⁴, and thereby braces of the same cardinality, for primes p greater than 5⁵. This thesis is based on three papers, each included in its entirety. The included content largely retains its original presentation, albeit with some changes to the sequence of section inclusions, and with occasional changes in wording, or minor corrections. The mathematical content of the referenced papers is unchanged
‘It’s not easy’: health and transnational belonging among West African immigrants in Philadelphia
No full textThis thesis explores embodied experiences of migration among West African (im)migrants in Philadelphia, Pennsylvania, USA. Southwest Philadelphia is home to numerous long-standing and continually growing diasporas from across West Africa, primarily Liberia, Sierra Leone, Guinea-Conakry, Nigeria, Burkina Faso, Mali, Mauritania, the Gambia, and Côte d’Ivoire. Among the oldest and most substantial diasporas are those from Liberia and Sierra Leone, following the initial arrival of refugees in the early 1990s due to their respective civil wars (Liberia from 1989-2003; Sierra Leone from 1991-2001). Drawing on one year of fieldwork, this thesis explores how immigrants navigate, utilise, and experience differing forms of care in different clinical spaces, within families, and among diasporic communities, and in turn how these experiences reconfigure notions of belonging for individuals and families between Philadelphia and West Africa. Moreover, I argue that ethnographic attention to affective and bodily experiences of health, and the ways in which they become entangled with borders, is integral to understanding the production of post-migration ideas of a “better life”.
This thesis outlines empirically what I term the fragmented and fragile healthcare network, enacted by a range of medical staff and aligned social service providers located at various clinical spaces and organisations throughout the city to assist West African immigrants to access various types of care. I argue that care within this network is a locus for expressing and articulating (dis)connection and belonging between West Africa and America in often contradictory and contested ways. In conjunction with biomedical interventions, this thesis explores the production of kinship relations across borders. Although families and communities are physically separated, bodies continue to mediate relationships on both sides of the Atlantic as acts of care for the living and the dead become crucial terrains for both reaffirming and producing new subjectivities. Thus, migrating to Philadelphia for the “American dream” becomes a project that is inherently fragmented and fragile, strung together through documents, medicine, friendships, pain, and community. This thesis contributes to Medical Anthropology, Migration Studies, and African Studies by highlighting the lived experiences of specific communities within Philadelphia that have been elided in ethnographic research, and showing how life is made and lived within America’s stratified moral economy of care
The development of a novel infection model to test CSF1-Fc as a potential therapeutic for early onset sepsis
Get PDFSepsis is a potentially fatal dysregulation of the host immune response to infection, involving stages of hyperinflammation that mediate organ damage and dysfunction. Early onset sepsis (EOS) is defined as onset within the first 72 hours after birth. It is caused by ingestion or inhalation of infected fluids either in utero or intrapartum. The only treatments currently available are antibiotics. However, with the rise in antibiotic resistance and long-term morbidities associated with antibiotic-induced dysbiosis, an alternative therapy is required. Our laboratory proposes colony stimulating factor-1 fusion protein (CSF1-Fc) as a potential host-based therapeutic, to clear the underlying infection and limit inflammation in EOS, as it has been shown to increase the maturity of macrophages. The aim of this project was to develop a novel neonatal infection model, to be utilised in future for the testing of potential therapeutics. This involved pipette-feeding bacteria to postnatal day 0 (P0) pups. Variations of this model involved administration of a mixed bacterial culture extracted from faeces that represents the maternal microbiome, or of a pure strain of Klebsiella pneumoniae to act as a clinically relevant model. Tested bacterial doses ranged from 104 – 108 colony forming units (CFU) over a 24, 48, 72 hour or 15 day infection duration. Bacterial colonisation was evaluated by plating the blood, spleen, small intestine and perianal swabs onto agar plates and counting colony forming units. Immune cell numbers were determined through haematoxylin and eosin (H&E) staining and anti-ionized calcium-binding adaptor molecule-1 (IBA1) immunohistochemistry of liver tissue sections. Contrary to our prediction, the faecal slurry and the selected strains of K. pneumoniae were unable to establish an infection and recover bacteria in a dose-dependent manner. Instead, increased virulence of the infecting agent, higher doses and a different infection timeline could be employed to improve success. Therefore, this thesis forms the basis for future investigations and informs on methodologies that require further optimisations, in order to achieve a dose-response neonatal infection model
Integrating forecasts based on machine learning and anthropometric data into Kenya’s National Drought Monitoring Authority
Get PDF‘Lightning talk’ presented by Susana Constenla-Villoslada (University of California, Berkeley) at the Jameel Observatory Community of Practice meeting, Addis Ababa, 13-14 May 202
Quantum groups and cohomological Donaldson-Thomas theory
No full textWe calculate the deformed and non-deformed cohomological Hall algebra (CoHA)
of the preprojective algebra for the case of cyclic quivers by studying the Kontsevich-
Soibelman CoHA and using tools from cohomological Donaldson-Thomas theory.
We show that for the cyclic quiver of length K, this algebra is the universal enveloping
algebra of the positive half of a certain extension of matrix differential
operators on C*, while its deformation gives a positive half of an explicit integral
form of Guay’s Affine Yangian ¨ Yₕ₁,ₕ₂(gl(K)). By the main theorems of [BD23]
and [SV23], we also determine the Maulik-Okounkov Yangian for the case of cyclic
quivers. Furthermore, we provide evidence for the strong rationality conjecture,
calculate the spherical subalgebra of the non-deformed CoHA for any quiver without
loops, the associated graded algebra of non deformed CoHA with respect to
perverse filtration for any quiver, recover results about the CoHA of compactly
supported semistable sheaves on the minimal resolution of the Kleinian singularity
C²/Zₖ, identify a commutative algebra inside the additive shuffle algebra
associated to the cyclic quiver and following the ideas of Davison-Hennecart-
Kinjo-Schiffmann-Vasserot, we prove existence of cocommutative coproduct on
the CoHA. We end by conjecturally relating the obtained integral form with the
algebra defined by Gaiotto-Rapčák-Zhou, in the context of twisted M-theory.
This thesis is based on [Jin24]
On the impact of super-massive black hole feedback on the properties of X-ray emitting gas in galaxy groups and clusters
No full textThe multiphase gas that exists in the ambient medium of large () cosmic halos is strongly impacted by feedback from a central Active Galactic Nuclei (AGN). This gas, which makes up either the Intra-group Medium (IGrM) or Intra-cluster Medium (ICM) depending on the size of the halo, determines the evolution of the central massive galaxy and of the system as a whole, through processes such as accretion onto galaxies, ram-pressure stripping of satellites, and the transfer of energy via pressure and gravity waves. Exactly how these processes operate, and how this gas is impacted itself by powerful kinetic feedback from the central black hole, remains poorly understood. Cosmological simulations, coupled with high-resolution zoom-ins, are ideal laboratories in which to investigate the feedback cycle as a whole and make new inroads into understanding the flow of baryons, energy, and entropy in these systems.
In this thesis I have started by constructing a powerful software pipeline to create and analyse highly-realistic mock X-ray observations of simulated halos in the Simba suite. Using this package, MOXHA (which is publicly available), I have made a mock predictive Athena-IFU catalogue of the most massive halos in Simba and used observer-like techniques to translate the simulation data to observation space as precisely as possible.
I find good agreement in Simba with observed X-ray scaling relations, reproducing an under-luminous, overly-hot mass range below . This is indicative of the black hole jet feedback more easily evacuating the low-entropy gas from the centers of these potential wells, and demonstrates the feedback model's ability to provide sufficient energy and momentum flux. Furthermore I demonstrated that the inferred halo mass bias is significantly under-estimated when mass-weighting the pressure profile in simulations. My work shows that for proper comparison to observation data it is imperative to use a luminosity-weighting scheme, which I find agrees very well with a full deprojection and spectral-fitting procedure.
Motivated by some promising features I identified in several of the mock observations presented in the aforementioned study that appeared to be X-ray cavities, I decided to investigate further using higher-resolution zoom-in simulations. This work makes up the next chapter of this thesis. Using a suite of 34 zooms spanning in , I found many viable X-ray bubbles over the timescale of Myr - the first time such features have been shown to be produced ad-hoc by a modern cosmological simulation. I found that these bubbles matched the morphology of observed cavities well. I furthermore found that the relation between the derived cavity enthalpy and the bolometric cooling luminosity also shows excellent agreement, and again shows an ability for the central AGN in the Simba model to strongly heat and reduce the entropy in the centers of these halos.
This thesis ends with a discussion of my current work, which is focused on the response of cold gas in these zoom-in simulations to the inflation and subsequent evolution of AGN-inflated X-ray cavities, and to the motion of satellite galaxies. In this work I present some preliminary results showing how cold gas can be precipitated from the stirring of the IGrM/CGM by kinetic feedback from supermassive black holes. I also investigate the impact of bubbles and satellites on the ratio, which determines whether gas can fully condense in these systems
Integrating biosemiotics with urban tactical navigation: a framework for novel spatial strategies
Get PDFAs cities expand, challenges such as housing shortages, inefficient land use, and environmental pressures demand innovative urban solutions. Traditional planning often overlooks underutilised urban spaces, or “gap sites”, which hold untapped potential for habitation. This research explores how biosemiotics and tactical urbanism can transform these spaces into semi-autonomous habitation pods, offering a data-driven, adaptable approach to urban development.
By conceptualising urban environments as communicative ecosystems, this study employs a semiotic lens to decode the interactions between people, objects, and spaces. Drawing on Peircean semiotics, the research frames habitation pods as objects, data practices as sign-vehicles, and urban citizens as interpretants who interact with and influence their surroundings. This methodology integrates economic mapping to assess land values, social media analytics to track urban engagement, and EEG-based neuroscientific data to evaluate the emotional impact of different environments.
The study reimagines architecture as an adaptive, responsive system, where habitation pods function as dynamic entities within the urban fabric, akin to biological organisms in a living ecosystem. Through geolocated data and behavioural analysis, the research demonstrates how semiotic-driven design can efficiently identify urban spaces suitable for habitation, offering scalable, citizen-centred strategies for urban adaptation.
By bridging semiotic theory with data-driven urban analysis, the study presents an alternative approach to spatial planning, shifting from static masterplans to flexible, real-time interventions. The findings contribute to contemporary debates on tactical urbanism, smart cities, and biosemiotics, offering insights for architects, urban planners, and policymakers in rethinking urban space as communicative, adaptable, and inclusive
Credentialisation of identification: how digital ID wallets shape digital identification, citizenship and power
Get PDFThis dissertation looks at the European Digital Identity (EUDI) Wallet: an app that will store digital versions of identity documents, as well as other information- ranging from social security numbers to supermarket loyalty cards - and aims to put citizens in control of these personal data. Specifically, the dissertation investigates the technological arrangement, Self-Sovereign Identity (SSI) technology, that underpins this wallet, illustrating that this set of technologies and ideas that was originally developed for the internet found its way to the heart of state-backed identification practices in the European Union. In light of this shift, the dissertation investigates how the wallet shapes digital citizenship, as well as the new power dynamics that it brings about. In so doing, it aims to elucidate the shifting sites of power in contemporary governance.
While the dissertation situates itself in two main sociological literatures, digital sociology and political sociology, it also borrows from science and technology studies. Drawing on interviews, field work at industry events and documentary research, the dissertation argues that the shift to a wallet-based identification system is not just technological; it also signifies an important societal change, which will have real implications for how identification is ‘done’, and how citizens gain access to services. The dissertation develops the concept of ‘credentialisation’ to analyse how the wallet turns different and previously separate parts of people’s lives into ‘verifiable’ information. As this information is subsequently brought together in one wallet, this signifies the collapse of different social institutions and roles into one technical system. This, the dissertation argues, has significant consequences, because 1.) it produces a new subject- a ‘user-citizen’ - who is supposed to be tech-savvy and responsible, and has to decide (to an extent) on the terms of their own datafication, and 2.) it instantiates new power dynamics around identification, as there is simultaneous competition over and collaboration between state and corporate actors concerning the design of digital ID wallets. One of the first in-depth studies on this topic, this dissertation contributes to the literature on digital identification and digital citizenship, demonstrating that state-issued identification is changing, as the credentialisation of identification gives rise to new technologies, new digital citizens, and new power dynamics
Access Repository Dashboard
Do you manage Edinburgh Research Archive? Access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard!