In massive-star binary systems, upon reaching later stages of stellar evolution one star can expand as a giant and envelope its companion in what is called a common envelope phase. The enveloped companion, here a neutron star, begins to accrete matter. The angular momentum of the accreting material results in the formation of an accretion disk. Accretion of hydrogen rich onto common-envelope-phase neutron stars can result in material ejected from the accretion disk having undergone burning near the neutron star’s surface [1]. Not much is understood about what nucleosynthesis occurs in this system. However, Keegans (2019) found that accreting neutron star common envelopes have the potential to impact galactic chemical evolution (GCE) [1].
Our preliminary results show that this astrophysical scenario can produce large amounts of light p-nuclides 92Mo, 96Ru and 98Ru – upwards of one order of magnitude more than their initial abundances in our simulations. This is significant as these isotopes are all underproduced in current p-process models and their origins are not known [2, 3].
The presented work builds on Keegans et al. (2019), which modelled accreting neutron star common envelopes without the inclusion of angular momentum, and Abrahams et al. (2023), which presented initial results on updated models which included the impact of angular momentum [1,4]. We will present yields from our common envelope simulations and discuss the nucleosynthesis which leads to high production of particular light p-nuclides.
[1] Keegans J., Fryer C.L., Jones S.W., Côte B., Belczynski K., Herwig F., Pignatari M., et al., 2019, MNRAS, 485, 620. doi:10.1093/mnras/stz368
[2] Roberti L., Pignatari M., Psaltis A., Sieverding A., Mohr P., Fulop Z., Lugaro M., 2023, A&A, 677, A22. doi:10.1051/0004-6361/202346556
[3] Role of Core-collapse Supernovae in Explaining Solar System Abundances of p Nuclides, C. Travaglio, T. Rauscher, A. Heger, M. Pignatari, C. West
[4] Abrahams S.E.D., Fryer C., Hall-Smith A., Laird A., Diget C., 2023, EPJWC, 279, 10002. doi:10.1051/epjconf/202327910002
The first stars can be constrained by the chemical composition of distant galaxies. It is crucial to understand how and when the first stars formed to understand the formation and evolution of our Universe. The latest observational data reveal unprecedented information about the chemical enrichment of the early Universe, which seems to behave differently from the local Universe. The first stars, being very massive, enrich their metal-poor environment in an uncertain way. To predict the abundances of the first galaxies, we include nucleosynthesis yields from Population III stars up to 300Msun, including faint supernovae, Wolf Rayet and Pair Instability Supernovae into our state-of-the-art hydrodynamical cosmological simulations. Our code (based on Gadget-3) also includes the latest nucleosynthesis yields from population II stars (from Kobayashi et al. 2020) for all stellar mass ranges. We predict the chemical abundance evolution of galaxies for different elements from the early Universe to the local Universe. For example, we find that the N/O abundance gives a systematically larger value with nucleosynthesis yields from Population III stars, which is comparable with observational data of the GN-z11 galaxy. I also discuss the evolution of metallicity gradients and elemental abundances of the intergalactic medium. We constrain our model by comparing it with observational data from the James Web Space Telescope (JWST) and the Atacama Large Millimeter/submillimeter Array (ALMA).
The properties of the first (Pop. III) stars remain a mystery. The chemistry of relic environments, enriched only by the supernovae of these first stars, offer an exciting avenue to study this population. Stellar relics are often found in the local Universe while gaseous relics probe the chemistry of low density structures at earlier epochs (z>2). I will discuss the complementary nature of these searches and how they can be used together to understand early chemical evolution and structure formation. Particularly, I will focus on the most metal-poor DLAs found at z~3 and the associated high-precision abundance determinations. This will include an updated view, provided by new data, on both the [O/Fe] enhancement seen at the lowest metallicities and the 12C/13C isotope ratio. Uniquely, this isotope ratio can be used to probe the existence of low-mass (i.e. 1M_sun) Pop III stars and the enrichment timescale of these near-pristine DLAs.
When asked to make therapeutic antimicrobial decisions on clinical rotations, veterinary students will often hesitate. Given that there are frequently several possible options and not always a defined “correct or incorrect” choice when it comes to antimicrobial selections, we sought to offer students more opportunities to practice decision making in a continuum. We developed a series of small animal case vignettes with a continuum of pharmacologic choices. Rather than making a discrete drug selection on a multiple-choice exam, students use a slider bar tool to indicate the relative safety and/or efficacy of the drug in question; they also provide a written justification for their selection. Slider bar selections and justifications are anonymized and downloaded into a corresponding file for instructor review. Student participation engenders real time feedback for instructors, which shapes key learning goals for class sessions and addresses misconceptions. An example of the case used to modify a teaching approach in the classroom highlights how the tool provides insight into students’ mindset. The case detailed an intact male dog with clinical signs of a urinary tract infection and supportive urinalysis data. Students evaluated the safety and efficacy of five different drugs. Most of the students deemed the aminopenicillin or potentiated aminopenicillin as safe and effective for use in this patient; most missed the signalment in the case and ignored the role of the prostate. The case sharpened the accompanying class discussion on the significance of antimicrobial selection in an intact male dog with a urinary tract infection.
In the pursuit of environmental sustainability, the development of energy-efficient artificial intelligence (AI) models becomes imperative. This work presents a novel approach to bioacoustic monitoring using self-supervised learning (SSL) of bird sound representations. We employ MobileNetV3, a lightweight deep learning model designed to focus on low computational cost while maintaining high performance. Our method leverages the capability of SSL to exploit unlabeled data, significantly reducing the dependency on extensive labelled datasets that is costly. By choosing appropriate data augmentation techniques to train MobileNetV3 in a self-supervised manner, we extract informative features from the data. These features achieve robust few-shot learning capabilities, enabling accurate bird species recognition with one labeled example per class. The application of such a lightweight model not only mitigates the environmental impact associated with the training of large-scale AI models but also enhances the feasibility of deploying AI solutions in resource-constrained environments. In future work, we aim to extend the application of our method to the large bird sound collection Xeno-Canto, which contains recordings from over 10,000 species. This scalability will allow us to learn more robust features, potentially enhancing the generalizability and effectiveness of our model across a broader spectrum of bioacoustic challenges. Additionally, by employing model compression techniques such as distillation, we can further reduce the model size, enabling the development of an ultra-compact model that is well-suited for deployment on small devices.
As the world transitions towards sustainable energy sources to combat climate change, wind power has emerged as a promising renewable alternative. The growing demand for wind energy generation has significant interest in developing accurate wind speed and energy forecasting models. Reliable forecasting is crucial for optimizing the planning and operation of wind power plants, enabling the seamless integration of this clean and sustainable energy source into the grid. While numerous models have been proposed in the past to predict wind speed and energy, their performance has been hindered by the inherent non-linearity and irregularity of wind patterns. To address this challenge and facilitate the widespread adoption of sustainable wind power, this research introduces a Novel Modular Red Deer Neural System (MRDNS) designed to effectively forecast wind speed and energy. The MRDNS harnesses data from wind turbine SCADA databases, which undergoes pre-processing to eliminate training flaws. Relevant features are then extracted, reducing the complexity of the prediction process. By analysing these features, the MRDNS employs a fitness function to predict wind speed and energy with enhanced accuracy, supporting the efficient utilization of this renewable resource. The model achieved remarkable performance, boasting a prediction accuracy of 99.99% for wind power forecasting, with an MSE of 0.0017 and an RMSE of 0.0422. For wind speed forecasting, the model yielded an MSE of 0.0003 and an RMSE of 0.0174.
Sensor-based remote healthcare monitoring offers a sustainable solution for detecting adverse health events in individuals with long-term conditions, directly in their homes. Traditional anomaly detection methods in noisy, multivariate real-world data often require large labelled datasets, complex AI models, extensive hyperparameter tuning, and frequent retraining to address data drift, limiting their scalability and explainability. Inspired by the simplicity and success of negative sample-free contrastive learning in computer vision, we propose a resource-efficient, self-supervised model that adapts to noise to improve anomaly detection. Our model has outperformed similar algorithms in detecting agitation and fall events in a real-world study of dementia patients. We enhanced model transparency through a ‘spatiotemporal attention map’ that pinpoints anomalies, fostering user trust and encouraging broader adoption. Our scalable, domain-agnostic solution can be applied across diverse healthcare, industrial, and urban environments, aligning with sustainable development goals, particularly in low-resource settings.
Convolutional neural networks (CNNs) have exhibited state-of-the-art performance in various audio classification tasks. However, their real-time deployment remains a challenge on resource-constrained devices such as embedded systems. In this paper, we analyze how the performance of large-scale pre-trained audio neural networks designed for audio pattern recognition changes when deployed on a hardware such as a Raspberry Pi. We empirically study the role of CPU temperature, microphone quality and audio signal volume on performance. Our experiments reveal that the continuous CPU usage results in an increased temperature that can trigger an automated slowdown mechanism in the Raspberry Pi, impacting inference latency. The quality of a microphone, specifically with affordable devices such as the Google AIY Voice Kit, and audio signal volume, all affect the system performance. In the course of our investigation, we encounter substantial complications linked to library compatibility and the unique processor architecture requirements of the Raspberry Pi, making the process less straightforward compared to conventional computers (PCs). Our observations, while presenting challenges, pave the way for future researchers to develop more compact machine learning models, design heat-dissipative hardware, and select appropriate microphones when AI models are deployed for real-time applications on edge devices.
Convolutional neural networks (CNNs) are commonplace in
high-performing solutions to many real-world problems, such as
audio classification. CNNs have many parameters and filters,
with some having a larger impact on the performance than others.
This means that networks may contain many unnecessary filters,
increasing a CNN’s computation and memory requirements while
providing limited performance benefits. To make CNNs more
efficient, we propose a pruning framework that eliminates filters
with the highest “commonality”. We measure this commonality
using the graph-theoretic concept of centrality. We hypothesise
that a filter with a high centrality should be eliminated as it
represents commonality and can be replaced by other filters without
affecting the performance of a network much. An experimental
evaluation of the proposed framework is performed on acoustic
scene classification and audio tagging. On the DCASE 2021 Task
1A baseline network, our proposed method reduces computations
per inference by 71% with 50% fewer parameters at less than a
two percentage point drop in accuracy compared to the original
network. For large-scale CNNs such as PANNs designed for audio
tagging, our method reduces 24% computations per inference with
41% fewer parameters at a slight improvement in performance.
Cells secrete extracellular vesicles (EVs) through various biogenesis pathways, resulting in distinct molecular compositions even when originating from the same cells. Analyzing individual EVs is challenging due to the need to overcome issues such as nanoscale size, heterogeneity of EVs, and measurement accuracy. Overcoming these challenges in EV research not only advances the field of biopsy but also enables progress in the biological research of EVs. In this study, we analyze quantifying membrane proteins using a novel methodology based on single-molecule fluorescence spectroscopy. We used TIRF based fluorescence imaging or Fluorescence Correlation Spectroscopy (FCS) to observe antibodies or aptamers binding to single EVs. Our results demonstrate the quantification of various membrane proteins, including CD63 and CD81. Consequently, we develop a membrane protein quantification assay for individual EVs using single-molecule and particle fluorescence imaging spectroscopy. Through this single molecule assay, we will elucidate the distribution of membrane proteins in EVs derived from various tumor cells, potentially enabling their use in future biopsies or quantification studies.