检索结果-内蒙古大学图书馆

Computational Intelligence Approaches and Applications (ICCIAA), International Conference on

作者： Mohammad Abu-Arqoub Abed Alkarim Banna Nuha El-Khalili Mohammad Al-Shaikh Hasan Department of Computer Science University of Petra Amman Jordan Data Science & Artificial Intelligence Dept. University of Petra Amman Jordan Software Engineering Department University of Petra Amman Jordan Computer Science Department Brunel University London London UK

ISBN: (数字)9798331523657

ISBN: (纸本)9798331523664

This article presents a comprehensive study on designing and implementing a dashboard for enhanced data visualization and query support at the University of Petra. The system leverages retrieval-augmented generation (RAG) and large language models (LLMs) to support diverse document types, including curricula, course descriptions, and program outcomes, alongside intended learning outcome (ILO)-related files. Our implementation demonstrates significant improvements in data accessibility and query response times while maintaining high accuracy in information retrieval and visualization. Through extensive evaluation, we show that this innovative approach transforms data management processes in higher education by enabling natural language interactions with educational data systems, building upon established business intelligence frameworks while introducing advanced AI capabilities.

关键词： Large language models Retrieval augmented generation Natural languages data visualization Transforms Educational technology Information retrieval data systems Business intelligence Time factors

Molecule Design by Latent Prompt Transformer 38

学校读者我要写书评

暂无评论

Molecule Design by Latent Prompt Transformer

38th Conference on Neural Information Processing Systems, NeurIPS 2024

作者： Kong, Deqian Huang, Yuhao Xie, Jianwen Honig, Edouardo Xu, Ming Xue, Shuanghong Lin, Pei Zhou, Sanping Zhong, Sheng Zheng, Nanning Wu, Ying Nian Department of Statistics and Data Science UCLA United States Institute of Artificial Intelligence and Robotics Xi'an Jiaotong University China BioMap Research China Akool Research United States Institute of Engineering in Medicine UCSD United States Shu Chien-Gene Lay Department of Bioengineering UCSD United States

This work explores the challenging problem of molecule design by framing it as a conditional generative modeling task, where target biological properties or desired chemical constraints serve as conditioning variables. We propose the Latent Prompt Transformer (LPT), a novel generative model comprising three components: (1) a latent vector with a learnable prior distribution modeled by a neural transformation of Gaussian white noise;(2) a molecule generation model based on a causal Transformer, which uses the latent vector as a prompt;and (3) a property prediction model that predicts a molecule's target properties and/or constraint values using the latent prompt. LPT can be learned by maximum likelihood estimation on molecule-property pairs. During property optimization, the latent prompt is inferred from target properties and constraints through posterior sampling and then used to guide the autoregressive molecule generation. After initial training on existing molecules and their properties, we adopt an online learning algorithm to progressively shift the model distribution towards regions that support desired target properties. Experiments demonstrate that LPT not only effectively discovers useful molecules across single-objective, multi-objective, and structure-constrained optimization tasks, but also exhibits strong sample efficiency. © 2024 Neural information processing systems foundation. All rights reserved.

关键词：

A Sensor Ranking Approach for Edge–Fog–Cloud Environments

学校读者我要写书评

暂无评论

SN computer science 2023年第4期4卷 392页

作者： Costa, Felipe S. Nassar, Silvia M. Dantas, Mario A. R. Department of Informatics and Statistics (INE) Federal University of Santa Catarina (UFSC) Florianopolis Brazil Department of Computer Science (DCC) Federal University of Juiz de Fora (UFJF) Juiz de Fora Brazil INESC P&D Brazil Santos Brazil

This research work presents a method for ranking sensors using the data produced by these devices. The method classifies the data, identifying the occurrence of failures in sensors and anomalies in the environments, aiming to maintain a reliability-based sensor ranking list. To generate the ranking list, overcoming the challenges implicit in this activity, the method adopts the theory of active perception as a basis. This approach divides the perception activity into levels that progressively add more sense to the information generated by the sensor, thus providing more reliability to the task of classifying the data generated by the sensors, without spending resources. This step allows you to create the ranking list in which the most reliable sensors will be at the top of the list. This list is managed through a distributed hash table to meet the distributed requirement of the Internet-of-Things (IoT) environment. The proposal was evaluated using four real data sets. The results of this research demonstrate that the proposed approach can provide high reliability in the use of sensor data, using low computational resources and, thus, reducing latency in the processes of selection and use of sensors. © 2023, The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd.

关键词： Active perception Anomaly Failures Latency Reliability Sensor ranking

Syrupy Mouthfeel and Hints of Chocolate -Predicting Coffee Review Scores using Text Based Sentiment

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Lohse, Christopher Lemsom, Jeroen Kalogiratos, Athanasios Department of Computer Science and Statistics Trinity College Dublin Dublin Ireland

This paper uses textual data contained in certified (q-graded) coffee reviews to predict corresponding scores on a scale from 0-100. By transforming this highly specialized and standardized textual data in a predictor space, we construct regression models which accurately capture the patterns in corresponding coffee bean scores. © 2023, CC BY-SA.

关键词： Metadata

Dynamical Conditional Optimal Transport through Simulation-Free Flows

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Kerrigan, Gavin Migliorini, Giosue Smyth, Padhraic Department of Computer Science University of California Irvine United States Department of Statistics University of California Irvine United States

We study the geometry of conditional optimal transport (COT) and prove a dynamical formulation which generalizes the Benamou-Brenier Theorem. Equipped with these tools, we propose a simulation-free flow-based method for conditional generative modeling. Our method couples an arbitrary source distribution to a specified target distribution through a triangular COT plan, and a conditional generative model is obtained by approximating the geodesic path of measures induced by this COT plan. Our theory and methods are applicable in infinite-dimensional settings, making them well suited for a wide class of Bayesian inverse problems. Empirically, we demonstrate that our method is competitive on several challenging conditional generation tasks, including an infinite-dimensional inverse problem. © 2024, CC BY.

关键词： Inverse problems

Disguise Face Classification Using EfficientNet Deep Learning 1st

学校读者我要写书评

暂无评论

Disguise Face Classification Using EfficientNet Deep Learnin...

1st International Conference on Human-Centric Smart Computing, ICHCSC 2022

作者： Padmashree, G. Wagle, Shruti G. Karunakar, A.K. Department of Data Science and Computer Applications Manipal Institute of Technology Manipal Academy of Higher Education Manipal India Department of Information and Communication Technology Manipal Institute of Technology Manipal Academy of Higher Education Manipal India

ISBN: (纸本)9789811954023

With the rise in popularity of social media and intelligent gadgets, one essential biometrics for identifying people is their face. The efficiency of existing automatic face recognition systems has decreased due to factors such as face ageing, conceals, and pose variations. Face recognition algorithms must be more accurate for recognizing faces hidden behind masks and makeup, as security and surveillance requirements become more stringent. Disguise face classification acts as a standalone early warning system in such scenarios. The major goal of this research was to see how the EfficientNet family of models compares to the current state-of-the-art architecture in terms of disguise face classification. For disguise face classification, the EfficientNet deep learning architecture was proposed in this paper. The models were trained and tested using the Disguise Faces in Wild (DFW) 2018 data set. The transfer learning method was used to train EfficientNet and other deep learning models. Results obtained proved that EfficientNet-B3 has outperformed other EfficientNet architectures with 92.2% precision and 93.9% accuracy. © 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

关键词： Deep learning

Feature-wise and Sample-wise Adaptive Transfer Learning for High-dimensional Linear Regression

学校读者我要写书评

暂无评论

arXiv 2024年

作者： He, Zelin Sun, Ying Liu, Jingyuan Li, Runze Department of Statistics Pennsylvania State University United States School of Electrical Engineering and Computer Science Pennsylvania State University United States MOE Key Laboratory of Econometrics Department of Statistics and Data Science School of Economics Wang Yanan Institute for Studies in Economics Fujian Key Lab of Statistics Xiamen University China

We consider the transfer learning problem in the high dimensional linear regression setting, where the feature dimension is larger than the sample size. To learn transferable information, which may vary across features or the source samples, we propose an adaptive transfer learning method that can detect and aggregate the feature-wise (F-AdaTrans) or sample-wise (S-AdaTrans) transferable structures. We achieve this by employing a fused-penalty, coupled with weights that can adapt according to the transferable structure. To choose the weight, we propose a theoretically informed, data-driven procedure, enabling F-AdaTrans to selectively fuse the transferable signals with the target while filtering out non-transferable signals, and S-AdaTrans to obtain the optimal combination of information transferred from each source sample. We show that, with appropriately chosen weights, F-AdaTrans achieves a convergence rate close to that of an oracle estimator with known transferable structure, and S-AdaTrans recovers existing near-minimax optimal rates as a special case. The effectiveness of the proposed method is validated using both synthetic and real data, demonstrating favorable performance compared to the existing methods. Copyright © 2024, The Authors. All rights reserved.

关键词： Transfer learning

The Fast and the Furious: Tracking the Effect of the Tomoa Skip on Speed Climbing

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Chou, Caleb Kaplan, Andee Department of Computer Science Colorado State University United States Department of Statistics Colorado State University United States

Sport climbing is an athletic discipline comprised of three sub-disciplines- lead climbing, bouldering, and speed climbing. These three sub-disciplineshave distinct goals, resulting in specialization of athletes into one of thethree events. The year 2020 marked the first inclusion of sport climbing in theOlympic Games. While this decision was met with excitement from the climbingcommunity, it was not without controversy. The International Olympic Committeehad allocated one set of medals for the entire sport, necessitating thecombination of sub-disciplines into one competition. As a result, athletes whospecialized in lead and bouldering were forced to train and compete in speedfor the first time in their careers. One such athlete was Tomoa Narasaki, aWorld Champion boulderer, who introduced a new method of approaching the speedevent. This approach, deemed the Tomoa Skip (TS), was subsequently adopted bymany of the top speed climbers. Concurrently, speed records fell rapidly (from5.48s in 2017 to 4.90s in 2023). Speed climbing involves ascending a 15m wallcontaining the same pattern of obstacles. Thus, records can be compared acrosstime. In this paper we investigate the effect of the TS on speed climbing byanswering two questions: (1) Did the TS result in a decrease in speed times?and (2) Do climbers who utilize the TS show less consistency? The success ofthe TS highlights the potential of collaboration between different disciplinesof sport, showing athletes of diverse backgrounds may contribute to theevolution of competition. Copyright © 2024, The Authors. All rights reserved.

关键词：

Probabilistic Rainfall Downscaling: Joint Generalized Neural Models with Censored Spatial Gaussian Copula

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Huk, David Adewoyin, Rilwan A. Dutta, Ritabrata Department of Statistics The University of Warwick United Kingdom Department of Computer Science and Engineering Southern University of Science and Technology China

This work introduces a novel approach for generating conditional probabilistic rainfall forecasts with temporal and spatial dependence. A two-step procedure is employed. Firstly, marginal location-specific distributions are jointly modelled. Secondly, a spatial dependency structure is learned to ensure spatial coherence among these distributions. To learn marginal distributions over rainfall values, we introduce joint generalised neural models which expand generalised linear models with a deep neural network to parameterise a distribution over the outcome space. To understand the spatial dependency structure of the data, a censored latent Gaussian copula model is presented and trained via scoring rules. Leveraging the underlying spatial structure, we construct a distance matrix between locations, transformed into a covariance matrix by a Gaussian Process Kernel depending on a small set of parameters. To estimate these parameters, we propose a general framework for the estimation of Gaussian copulas employing scoring rules as a measure of divergence between distributions. Uniting our two contributions, namely the joint generalised neural model and the censored latent Gaussian copulas into a single model, our probabilistic approach generates forecasts on short to long-term durations, suitable for locations outside the training set. We demonstrate its efficacy using a large UK rainfall data set, outperforming existing methods. © 2023, CC BY.

关键词： Location

Secure UPI: Machine Learning-Driven Fraud Detection System for UPI Transactions

学校读者我要写书评

暂无评论

Secure UPI: Machine Learning-Driven Fraud Detection System f...

Disruptive Technologies (ICDT), International Conference on

作者： Rupa Rani Adnan Alam Abdul Javed Department of Computer Science & Engineering Ajay Kumar Garg Engineering College Ghaziabad Ghaziabad U.P. India Department of Computer Science & Engineering Banasthali Vidyapith Rajasthan Department of Computer Science & Engineering-Data Science ABES Institute of Technology Ghaziabad U.P. India

Secure UPI specializes in developing an advanced fraud detection gadget the usage of the effective XGBoost device getting to know set of rules to create an advanced fraud identity device. XGBoost is a properly-proper alternative for enhancing the precision of fraud detection fashions because of its reputation for dealing with tricky datasets with efficiency and its music record of success throughout multiple industries. In order to extract pertinent records, like transaction quantity, frequency, and place, our approach preprocesses UPI transaction facts. This article makes use of a labelled dataset to train the XGBoost version in order that we may also take gain of its sturdy prediction talents and capacity to handle imbalanced datasets. To help create a system that is less difficult to apprehend and use, feature importance evaluation is used to discover essential symptoms of feasible fraud. After training, the model is covered right into a real-time UPI transaction tracking device, in which it maintains an eye fixed out for any suspicious traits in incoming transactions. In order to lessen the results of fraudulent activity, the system is constructed with 98.2 % accuracy to send out instant notifications and take preventive steps. This challenge allows in improving UPI transaction security and advancing economic era are accomplished through demonstrating the performance of machine learning in fraud detection.

关键词：