搜索结果：25000

The mass independent energy reconstruction of cosmic rays is crucial for understanding their origin, acceleration, and propagation. Precise measurement of the primary energy can also lead to better mass classification and could enable energy dependent anisotropy maps for individual elements. The GRAPES-3 experiment located in Ooty consisting of 400 scintillator detector array placed 8 m apart covering an area of 25000 m$^2$ with a dedicated muon detector made of 3712 proportional counters, is designed to do these kinds of measurements. Previously electron size calibration curves have been used to find primary energy in the GRAPES-3 data analysis framework however significantly better precision can be established using graph neural network. Thus, in this work we have implemented a modular and dynamic GNN based reconstruction algorithm that automates feature mapping. We demonstrate how the model is learning by studying its latent space and show that scaling the metric in the latent space can lead to further improvements in response resolution. Fine-tuned strategies are presented and a thorough comparison of the reconstructed energy and bias is done for different fine-tuned models alo

The photochemical CO2 reduction reaction represents a zero-carbon pathway for converting CO2 into value-added chemicals, yet its industrial implementation has been constrained by low selectivity and product diversity. Dirac nodal arc semimetals characterized by ultrahigh carrier mobility with over 25000 cm2 V-1 s-1 offer a promising platform to search for efficient catalysts for CO2 conversion. Herein, we demonstrate that strategic Pt incorporation into PdSn4 optimizes the electronic structure and carrier dynamics of this Dirac semimetal. Experimental and theoretical analyses reveal that the resulting Pd-Sn-Pt local electronic structure redistributes charge density around Pd and Pt atoms, which facilitates C-C coupling via *OC-COH and *OC-CHOH intermediates and enhances carrier mobility by 40% versus the pristine PdSn4 single crystal. The optimized Pd0.4Pt0.6Sn4 single crystal achieves C2H4 with formation rate of 0.000328 mol g-1 h-1, product selectivity of 73.1% and electron-based selectivity of 89%. This work establishes electronic-structure-tunable Dirac semimetals as a new paradigm for multi-carbon photochemical CO2 reduction, providing a design strategy for next-generation pho

Quadratic Unconstrained Binary Optimization (QUBO) is a versatile framework for modeling combinatorial optimization problems. This study benchmarks five software-based QUBO solvers: Neal, PyTorch (CPU), PyTorch (GPU), JAX, and SciPy, on randomly generated QUBO matrices ranging from 1000x1000 to 45000x45000, under six convergence thresholds from 10^-1 to 10^-6. We evaluate their performance in terms of solution quality (energy) and computational time. Among the solvers tested, Neal achieved the lowest energy values but was limited to problems with up to 6000 variables due to high memory consumption. PyTorch produced slightly higher energy results than Neal but demonstrated superior scalability, solving instances with up to 45000 variables. Its support for GPU acceleration and CPU multi-threading also resulted in significantly shorter runtimes. JAX yielded energy values slightly above those of PyTorch and was limited to 25000 variables, with runtimes comparable to PyTorch on GPU. SciPy was the most constrained solver, handling only up to 6000 variables and consistently producing the highest energy values with the longest computation times. These findings highlight trade-offs between

Television networks face high financial risk when making programming decisions, often relying on limited historical data to forecast episodic viewership. This study introduces a machine learning framework that integrates natural language processing (NLP) features from over 25000 television episodes with traditional viewership data to enhance predictive accuracy. By extracting emotional tone, cognitive complexity, and narrative structure from episode dialogue, we evaluate forecasting performance using SARIMAX, rolling XGBoost, and feature selection models. While prior viewership remains a strong baseline predictor, NLP features contribute meaningful improvements for some series. We also introduce a similarity scoring method based on Euclidean distance between aggregate dialogue vectors to compare shows by content. Tested across diverse genres, including Better Call Saul and Abbott Elementary, our framework reveals genre-specific performance and offers interpretable metrics for writers, executives, and marketers seeking data-driven insight into audience behavior.

Understanding emotional nuances in everyday language is crucial for computational linguistics and emotion research. While traditional lexicon-based tools like LIWC and Pattern have served as foundational instruments, Large Language Models (LLMs) promise enhanced context understanding. We evaluated three Dutch-specific LLMs (ChocoLlama-8B-Instruct, Reynaerde-7B-chat, and GEITje-7B-ultra) against LIWC and Pattern for valence prediction in Flemish, a low-resource language variant. Our dataset comprised approximately 25000 spontaneous textual responses from 102 Dutch-speaking participants, each providing narratives about their current experiences with self-assessed valence ratings (-50 to +50). Surprisingly, despite architectural advancements, the Dutch-tuned LLMs underperformed compared to traditional methods, with Pattern showing superior performance. These findings challenge assumptions about LLM superiority in sentiment analysis tasks and highlight the complexity of capturing emotional valence in spontaneous, real-world narratives. Our results underscore the need for developing culturally and linguistically tailored evaluation frameworks for low-resource language variants, while qu

We investigated the complete thermodynamic cycle of aluminium nanoparticles through classical molecular dynamics simulations, spanning a wide size range from 200 atoms to 11000 atoms. The aluminium-aluminium interactions are modelled using a newly developed Bayesian Force Field (BFF) from the FLARE suite, a cutting-edge tool in our field. We discuss the database requirements to include melted nanodroplets to avoid unphysical behaviour at the phase transition. Our study provides a comprehensive understanding of structural stability up to sizes as large as $3~ 10^5$ atoms. The developed Al-BFF predicts an icosahedral stability range of up to 2000 atoms, approximately 2 nm, followed by a region of stability for decahedra, up to 25000 atoms. Beyond this size, the expected structure favours face-centred cubic (FCC) shapes. At a fixed heating/cooling rate of 100K/ns, we consistently observe a hysteresis loop, where the melting temperatures are higher than those associated with solidification. The annealing of a liquid droplet further stabilizes icosahedral structures, extending their stability range to 5000 atoms. Using a hierarchical k-means clustering, we find no evidence of surface me

Model inversion and membership inference attacks aim to reconstruct and verify the data which a model was trained on. However, they are not guaranteed to find all training samples as they do not know the size of the training set. In this paper, we introduce a new task: dataset size recovery, that aims to determine the number of samples used to train a model, directly from its weights. We then propose DSiRe, a method for recovering the number of images used to fine-tune a model, in the common case where fine-tuning uses LoRA. We discover that both the norm and the spectrum of the LoRA matrices are closely linked to the fine-tuning dataset size; we leverage this finding to propose a simple yet effective prediction algorithm. To evaluate dataset size recovery of LoRA weights, we develop and release a new benchmark, LoRA-WiSE, consisting of over 25000 weight snapshots from more than 2000 diverse LoRA fine-tuned models. Our best classifier can predict the number of fine-tuning images with a mean absolute error of 0.36 images, establishing the feasibility of this attack.

We generated 25000 conversations labeled with Big Five Personality traits using prompt programming at GPT-3. Then we train Big Five classification models with these data and evaluate them with 2500 data from generated dialogues and real conversational datasets labeled in Big Five by human annotators. The results indicated that this approach is promising for creating effective training data. We then compare the performance by different training approaches and models. Our results suggest that using Adapter-Transformers and transfer learning from pre-trained RoBERTa sentiment analysis model will perform best with the generated data. Our best model obtained an accuracy of 0.71 in generated data and 0.65 in real datasets. Finally, we discuss this approach's potential limitations and confidence metric.

Timeseries classification as stochastic (noise-like) or non-stochastic (structured), helps understand the underlying dynamics, in several domains. Here we propose a two-legged matrix decomposition-based algorithm utilizing two complementary techniques for classification. In Singular Value Decomposition (SVD) based analysis leg, we perform topological analysis (Betti numbers) on singular vectors containing temporal information, leading to SVD-label. Parallely, temporal-ordering agnostic Principal Component Analysis (PCA) is performed, and the proposed PCA-derived features are computed. These features, extracted from synthetic timeseries of the two labels, are observed to map the timeseries to a linearly separable feature space. Support Vector Machine (SVM) is used to produce PCA-label. The proposed methods have been applied to synthetic data, comprising 41 realisations of white-noise, pink-noise (stochastic), Logistic-map at growth-rate 4 and Lorentz-system (non-stochastic), as proof-of-concept. Proposed algorithm is applied on astronomical data: 12 temporal-classes of timeseries of black hole GRS 1915+105, obtained from RXTE satellite with average length 25000. For a given timeseri

We have calculated a grid of massive star wind models and mass-loss rates for a wide range of metal abundances between 1/100 and 10 Z/Zsun. The calculation of this grid completes the Vink et al. (2000) mass-loss recipe with an additional parameter Z. We have found that the exponent of the power law dependence of mass loss vs. metallicity is constant in the range between 1/30 and 3 Z/Zsun. The mass-loss rate scales as Mdot \propto Z^0.85 Vinf^p with p = -1.23 for stars with Teff \ga 25000 K, and p = -1.60 for the B supergiants with Teff \la 25000 K. Taking also into account the metallicity dependence of Vinf, using the power law dependence Vinf \propto Z^0.13 from Leitherer et al. (1992), the overall result of mass loss as a function of metallicity can be represented by Mdot \propto Z^0.69 for stars with Teff \ga 25000 K, and Mdot \propto Z^0.64 for B supergiants with Teff \la 25000 K. Our mass-loss predictions are successful in explaining the observed mass-loss rates for Galactic and Small Magellanic Cloud O-type stars, as well as in predicting the observed Galactic bi-stability jump. Hence, we believe that our predictions are reliable and suggest that our mass-loss recipe be used

Despite its apparently simple nature with four valence electrons, the strontium dimer constitutes a challenge for modern electronic structure theory. Here we focus on excited electronic states of Sr$_2$, which we investigate theoretically up to 25000 cm$^{-1}$ above the ground state, to guide and explain new spectroscopic measurements. In particular, we focus on potential energy curves for the $1^1Σ^{+}_{u}$, $2^1Σ^{+}_{u}$, $1^1Π_{u}$, $2^1Π_{u}$, and $1^1Δ_{u}$ states computed using several variants of advanced \textit{ab initio} methods to benchmark them. In addition, a new experimental study of the excited $2^1Σ^{+}_{u}$ state using polarisation labelling spectroscopy is presented, which extends knowledge of this state to high vibrational levels, where perturbation by higher electronic states is observed. The available experimental observations are compared with the theoretical predictions and help to assess the accuracy and limitations of employed theoretical models. The present results pave the way for future more accurate theoretical and experimental spectroscopic studies.

A reliable determination of the basic physical properties and variability patterns of hot emission-line stars is important for understanding the Be phenomenon and ultimately, the evolutionary stage of Be stars. This study is devoted to one of the most remarkable Be stars, V1294 Aql = HD 184279. We collected and analysed spectroscopic and photometric observations covering a time interval of about 25000 d (68 yr). We present evidence that the object is a single-line 192.9 d spectroscopic binary and estimate that the secondary probably is a hot compact object with a mass of about 1.1-1.2 solar masses. We found and documented very complicated orbital and long-term spectral, light, and colour variations, which must arise from a combination of several distinct variability patterns. Attempts at modelling them are planned for a follow-up study. We place the time behaviour of V1294 Aql into context with variations known for some other systematically studied Be stars and discuss the current ideas about the nature of the Be phenomenon.

This paper describes an experimental investigation, by means of hot-wire anemometry, of the characteristics of velocity and temperature in a rotating turbulent boundary layer under isothermal and non-isothermal conditions. The ranges of experimental parameters are: Reynolds number from 10000 to 25000, rotational speed from 0 to 150 rpm, and y+ from 1.8 to 100. The relative temperature difference is held constant at 0.1. Detailed velocity and temperature distributions in the boundary layer are measured in the rotating state, and a new criterion for boundary layer segmentation under rotation is proposed. The applicability of boundary layer theory under the rotating state is extended. The influence of Coriolis force and buoyancy on the velocity and temperature distributions in the turbulent boundary layers are analyzed. Coriolis force is found to play an important role in the behavior of the boundary layer under rotation, as it shifts the velocity and temperature boundary layers. Under isothermal conditions, such effects can be classified according to the dominant force: viscous, Coriolis, or inertial. Under non-isothermal conditions, buoyancy occurs. The buoyancy induced by the Corio

Social media platforms are used by a large number of people prominently to express their thoughts and opinions. However, these platforms have contributed to a substantial amount of hateful and abusive content as well. Therefore, it is important to curb the spread of hate speech on these platforms. In India, Marathi is one of the most popular languages used by a wide audience. In this work, we present L3Cube-MahaHate, the first major Hate Speech Dataset in Marathi. The dataset is curated from Twitter, annotated manually. Our dataset consists of over 25000 distinct tweets labeled into four major classes i.e hate, offensive, profane, and not. We present the approaches used for collecting and annotating the data and the challenges faced during the process. Finally, we present baseline classification results using deep learning models based on CNN, LSTM, and Transformers. We explore mono-lingual and multi-lingual variants of BERT like MahaBERT, IndicBERT, mBERT, and xlm-RoBERTa and show that mono-lingual models perform better than their multi-lingual counterparts. The MahaBERT model provides the best results on L3Cube-MahaHate Corpus. The data and models are available at https://github.

In this study, we focus on learning Hamiltonian systems, which involves predicting the coordinate (q) and momentum (p) variables generated by a symplectic mapping. Based on Chen & Tao (2021), the symplectic mapping is represented by a generating function. To extend the prediction time period, we develop a new learning scheme by splitting the time series (q_i, p_i) into several partitions. We then train a large-step neural network (LSNN) to approximate the generating function between the first partition (i.e. the initial condition) and each one of the remaining partitions. This partition approach makes our LSNN effectively suppress the accumulative error when predicting the system evolution. Then we train the LSNN to learn the motions of the 2:3 resonant Kuiper belt objects for a long time period of 25000 yr. The results show that there are two significant improvements over the neural network constructed in our previous work (Li et al. 2022): (1) the conservation of the Jacobi integral, and (2) the highly accurate predictions of the orbital evolution. Overall, we propose that the designed LSNN has the potential to considerably improve predictions of the long-term evolution of mo

Many nuclear reactions used to create radioactive isotopes for nuclear research produce, in addition to the isotope of interest, many contaminants, which are often produced in much larger amounts than the isotope of interest. Many installations using the ISOL approach are therefore equipped with high-resolution mass separators to remove at least isotopes with a different mass number. In the present paper, we present the results of the commissioning of the DESIR HRS presently under development at LP2I Bordeaux (formerly CENBG). Optical aberrations are corrected up to 3rd order and a mass resolution of M/$Δ$M of 25000 is reached with a transmission of about 70% for a 133Cs+ beam at 25 keV.