28579 research outputs found
Sort by
Financial forecasting with the combination of physical and event-based time using genetic programming
Get PDFThis thesis explores the application of genetic programming (GP) within the directional changes (DC) framework for algorithmic trading. Traditional algorithmic trading methods rely on datasets with fixed time intervals, such as hourly or daily data, leading to a discontinuous representation of time. DC provides an alternative by transforming these datasets into event-driven sequences, allowing for a unique price analysis approach. The first part of the thesis compares GP with machine learning (ML) algorithms in algorithmic trading, focusing on factors like market data, time periods, forecasting windows, and transaction costs—variables often neglected in previous studies. A comprehensive evaluation of a GP-based financial approach is conducted, comparing it to nine popular ML algorithms and the buy-and-hold strategy, using daily data from 220 datasets across 10 international markets. Results show that GP not only yields profitable results but also outperforms ML algorithms in terms of risk and Sharpe ratio. The second part investigates GP within the DC framework, introducing two novel algorithms: GP-DC, which uses only DC-based indicators, and GP-DC-PT, which combines DC-based and physical-time indicators from technical analysis. Both approaches outperform non-DC-based GP strategies, technical analysis, and buy-and-hold benchmarks, with GP-DC-PT achieving an average return of over 18%, highlighting the advantage of incorporating DC into trading strategies. Finally, the thesis introduces two multi-objective optimization algorithms, MOO2 and MOO3, based on the NSGA-II framework, which optimize two and three fitness functions, respectively, using DC and physical-time indicators. Both MOO2 and MOO3 outperform single-objective methods, with MOO3 showing consistent improvements across all metrics. These findings suggest that incorporating directional changes significantly enhances trading strategies' return and risk performance
Tensor cross interpolation for global discrete optimization with application to Bayesian network inference
Get PDFGlobal discrete optimization is notoriously difficult due to the lack of gradient information and the curse of dimensionality, making exhaustive search infeasible.
Tensor cross approximation is an efficient technique to approximate multivariate tensors (and discretized functions) by tensor product decompositions based on a small number of tensor elements, evaluated on adaptively selected fibers of the tensor, that intersect on submatrices of (nearly) maximum volume.
The submatrices of maximum volume are empirically known to contain large elements, hence the entries selected for cross interpolation can also be good candidates for the globally maximal element within the tensor.
In this paper we consider evolution of epidemics on networks, and infer the contact network from observations of network nodal states over time.
By numerical experiments we demonstrate that the contact network can be inferred accurately by finding the global maximum of the likelihood using tensor cross interpolation.
The proposed tensor product approach is flexible and can be applied to global discrete optimization for other problems, e.g. discrete hyperparameter tuning
Self-fashioning, Food, and Masculinity in George III’s Monarchy
Get PDFGeorge III was a family man, a modest eater, and a thoughtful ruler who wrote about the big questions of the day, from royal sovereignty to the best methods of agriculture to feed a modern nation. His writings provide a glimpse of his version of monarchy, which placed him at the head of a national family, where he embodied the habits of self-regulation and temperance in keeping with the sensibilities of late eighteenth-century manhood. This article brings together George’s meals and his essays, considering the histories of food, masculinity, and self-fashioning, to argue that George was a monarch who embodied a new form of masculinity, as marked by his agricultural interests and insistence on a modest diet. His eating habits, along with his intellectual interests and public persona, bring us to the intersection between the private man and the public monarch. Drawing on newly digitised data, alongside contemporary caricatures and descriptions, and George’s own writing, we argue that moderation was central to George’s creation of an image that appealed to the emerging British nation of the late eighteenth century; food was central to this image, highlighting both his masculine self-control and his ability to be useful to the nation
Enhancing Forensic Audio Transcription with Neural Network-Based Speaker Diarization and Gender Classification
Get PDFForensic audio transcription is often compromised by low-quality recordings, where indistinct speech can hinder the accuracy of conventional Automatic Speech Recognition (ASR) systems. This study addresses this limitation by developing a machine learning-based approach to improve speaker diarization, a process critical for distinguishing between speakers in sensitive audio data. Previous research highlights the inadequacy of traditional ASR in forensic settings, particularly where audio quality is poor and speaker overlap is common. This paper presents a neural network specifically designed for gender classification, using 20 key acoustic features extracted from real forensic audio data. The model architecture includes input, hidden, and output layers tailored to differentiate male and female voices, with dropout regularization to prevent overfitting and hyperparameter optimization ensuring robust generalization across test data. The neural network achieved an average recall of 86.81%, F1 score of 85.67%, precision of 87.95%, and accuracy of 86.83% across varied audio conditions. This model significantly improves transcription accuracy, reducing errors in legal contexts and supporting judicial processes with more reliable, interpretable evidence from sensitive audio data
Mapping urban heat islands in Pune, India: ecological impacts and environmental challenges
Get PDFHeat waves increasingly affect cities, amplifying the urban heat island (UHI) effect, often measured through land surface temperature (LST). In Pune, India, rapid urbanization between 2013 and 2022 has driven significant land use and land cover changes, with a staggering 89.24% increase in built-up areas and a decline of 991.4 km2² in vegetation cover. Using satellite remote sensing data processed via Google Earth Engine, this study reveals a pronounced rise in LST, with mean temperatures increasing from 27°C in 2013 to 36°C by 2022, and a notable expansion in regions experiencing temperatures between 25°C and 32°C. Additionally, NO₂ levels slightly rose, further stressing environmental conditions. Central Pune was identified as a high-risk zone for adverse climatic impacts, emphasizing the urgent need for ecological conservation, climate adaptation, and sustainable urban planning to mitigate the growing UHI effect amidst accelerating urbanization in Indian cities
Does peacekeeping mitigate the impact of aid on conflict? Peacekeeping, humanitarian aid and violence against civilians
Get PDFPeacekeeping has been found to be effective in containing conflict and civilian victimization, while the findings for the effect of aid on violence are indeterminate. So far the effects of peacekeeping and aid on violence have mainly been studied separately, this article investigates, at the subnational level, the effect of humanitarian aid on one-sided violence conditional on the deployment of peacekeeping forces. Although humanitarian aid can occasionally exacerbate violence, it is argued that peacekeepers reverse this unintended consequence of the provision of aid. We argue that they do so by means of sharing information and the provision of security bubbles. Empirically, we look at the coincidence of subnational location of humanitarian agencies and peacekeeping troops and find support for the idea that the effect of aid on violence against civilians is conditional on the presence of peacekeepers
Exploring Pupil Dilation as an Indicator of Performance in Gaze-Based Robot Navigation for Assistive Technology
Get PDFHuman-robot interaction (HRI) based assistive devices play a crucial role for individuals with severe disability, significantly impacting their quality of life. A pivotal step towards creating a more human-centric HRI involves gaining a thorough understanding of the user's mental load such as cognitive load, stress, and fatigue, which can influence the performance of the system. Previous studies have found pupil dilation as a potential candidate for exploring mental workload. This paper explores the impact of pupil diameter variation on performance during an eye-tracking-based robot navigation task. Nineteen healthy individuals participated in the experiment where they used eye-gaze to activate different navigational buttons on a computer screen to control the movement of a mobile robot on a predefined trajectory for two rounds. The variation of pupil diameter is correlated to various performance parameters such as lap completion time and number of commands. Results show that the difference between the Gaussian means of the pupil diameter distribution during round1 and round2 is significantly correlated (rho =0.5, p-value =0.03) with the lap completion time while the correlation with the number of commands is also found to be strong (r h o=0.45, p-value =0.05). These quantifications of pupil diameter variations with performance measures have the potential to play a vital role in advancing the HRI systems as they can be used to predict the performance variation in real-time so that the HRI can be more responsive to the user's changing mental states, a key requirement for the practical usability and acceptability of such systems as assistiv
Exchange rate and commodity prices
Get PDFChapter 1 discusses how fluctuations in crude oil prices deeply influence the global economy, particularly in oil-exporting countries. Chapter 2 explores the economic dynamics of crude oil prices and their impact on the Real Effective Exchange Rate (REER) in oil-exporting countries. The analysis covers a diverse group of oil-exporting countries. We determine the REER based on the Behavioural Equilibrium Exchange Rate (BEER) model for these countries. Employing both time-series and panel methods, such as Johansen cointegration and the Autoregressive Distributed Lag (ARDL) bounds testing, the study investigates the real oil price on REER. While timeseries results are mixed, Pooled Mean Group (PMG) methods reveal a significant positive relationship between real oil price, Net Foreign Assets (NFA), and six-sector value-added deflator (6SECT) with REER. Chapter 3 examines whether exchange rates can reliably forecast international oil prices for oil-exporting countries. We focus on forecasting models for crude oil and its derivatives. We evaluate both in-sample and out-of-sample performance across global oil futures and major benchmarks (Brent, WTI, Dubai), using nominal exchange rates from Brazil, Canada, Colombia, Indonesia, Mexico, and Norway. Our findings reveal that the exchange rates of Brazil, Colombia, Mexico, and Norway possess substantial short-term forecasting power for crude oil and its derivative financial products, likely due to the significance of oil in their economies. In contrast, Canada and Indonesia have relatively small oil rents, which may explain why their exchange rates are less effective at predicting oil prices. Chapter 4 further investigates whether crude oil price predictability exists within a small window. We find that most of these windows occur during economic downturns, suggesting that oil-exporting countries’ exchange rates may gain predictive power in times of crisis. Chapter 5 concludes this thesis by highlighting key findings, addressing limitations, and suggesting avenues for future research
Linguistic synesthesia and embodiment: A study based on Mandarin modality exclusivity norms
No full textThis study aims to resolve the ongoing debate about sensory modality embodiment found in linguistic synesthesia by proposing an empirical model: Perceived Strength of Embodiment (PSE). The perceived strength of embodiment for sensory adjectives is measured based on the sensory ratings of the adjectives in the five sensory modalities, while the perceived strength of embodiment for each sensory modality is calculated based on the PSE of all adjectives according to their dominant modalities. PSE is designed to address a salient dilemma in the widely-accepted modality-based embodiment asymmetry: that is, such asymmetry fails to predict the directionality behaviors between sensory words because each sensory word is typically associated with more than one modality, and each may have different strengths of association. Based on an analysis of sensory adjectives, we find that a lexical concept-based embodiment asymmetry better explains the data than a modality-based embodiment asymmetry and, additionally, the lexical concept-based account is supported by Mandarin synesthetic compound adjective data. In sum, this paper argues that the PSE model is an empirical approach to measuring the degree of embodiment which furthers the understanding of the role of embodiment in the linguistic conceptualization of sensory perceptions
Prompt4LJP: prompt learning for legal judgment prediction
No full textThe task of legal judgment prediction (LJP) involves predicting court decisions based on the facts of the case, including identifying the applicable law article, the charge, and the term of penalty. While neural methods have made significant strides in this area, they often fail to fully harness the rich semantic potential of language models (LMs). Prompt learning is a novel paradigm in natural language processing (NLP) that reformulates downstream tasks into cloze-style or prefix-style prediction challenges by utilizing specialized prompt templates. This paradigm shows significant potential across various NLP domains, including short text classification. However, the dynamic word lengths of LJP labels present a challenge to the general prompt templates designed for single-word [MASK] tokens commonly used in many NLP tasks. To address this gap, we introduce the Prompt4LJP framework, a new method based on the prompt learning paradigm for the complex LJP task. Our framework employs a dual-slot prompt template in conjunction with a correlation scoring mechanism to maximize the utility of LMs without requiring additional resources or complex tokenization schemes. Specifically, the dual-slot template consists of two distinct slots: one dedicated to factual descriptions and the other to labels. This approach effectively tackles the challenge of dynamic word lengths in LJP labels, reformulating the LJP classification task as an evaluation of the applicability of each label. By incorporating a correlation scoring mechanism, we can identify the final result label. The experimental results show that our Prompt4LJP method, whether using discrete or continuous templates, outperforms baseline methods, particularly in charges and terms of penalty prediction. Compared to the best baseline model EPM, Prompt4LJP shows F1-score improvements of 2.25% and 4.76% (charge prediction and term of penalty prediction) with discrete templates, and 3.24% and 4.05% with the continuous template, demonstrating prompt4LJP ability to leverage pretrained knowledge and adapt flexibly to specific tasks. The source code can be obtained from https://github.com/huangqiongyannn/Prompt4LJP