The development of beyond-fifth-generation (B5G) communication systems introduces challenges in maintaining timing and frequency synchronization, especially in low SNR and extended coverage scenarios. Accurate carrier...
详细信息
ISBN:
(纸本)9798350361261;9798350361278
The development of beyond-fifth-generation (B5G) communication systems introduces challenges in maintaining timing and frequency synchronization, especially in low SNR and extended coverage scenarios. Accurate carrier frequency offset (CFO) estimation is crucial for establishing calls under such conditions. Existing methods, like maximum likelihood estimation, have limitations, while machinelearning (ML) techniques have shown promise in wireless communication. In this work, we propose an ML-based approach using Long Short-Term Memory (LSTM) neural networks and automated machinelearning (AutoML) to tune hyperparameters and improve CFO estimation accuracy. We compare our model with a gradient-boosting machine (GBM) approach and demonstrate superior accuracy. Our research addresses CFO estimation challenges in B5G systems and offers valuable insights for the development of robust techniques in advanced communication systems.
We propose a federated learning framework designed to effectively utilize diverse GPU architectures in edge devices. It offers wider compatibility with various GPU platforms than conventional approaches which tend to ...
详细信息
ISBN:
(纸本)9798350376975;9798350376968
We propose a federated learning framework designed to effectively utilize diverse GPU architectures in edge devices. It offers wider compatibility with various GPU platforms than conventional approaches which tend to focus on CUDA-based systems for enhancing the feasibility and cost-efficiency of machinelearning in edge computing scenarios. Our framework is structured for straightforward deployment, addressing the complexities of initial federated learning project setups. We also built tensor manipulation library from scratch as the core of this framework. This contribution extends the reach of federated learning, enabling more flexible and inclusive approaches in distributed machinelearning environments.
The proceedings contain 13 papers. The topics discussed include: large-scale clustering using MPI-based canopy;automated labeling of electron microscopy images using deep learning;large minibatch training on supercomp...
ISBN:
(纸本)9781728101804
The proceedings contain 13 papers. The topics discussed include: large-scale clustering using MPI-based canopy;automated labeling of electron microscopy images using deep learning;large minibatch training on supercomputers with improved accuracy and reduced time to train;communication-efficient parallelization strategy for deep convolutional neural network training;auto-tuning TensorFlow threading model for CPU backend;training speech recognition models on HPC infrastructure;automated parallel data processing engine with application to large-scale feature extraction;aluminum: an asynchronous, GPU-aware communication library optimized for large-scale training of deep neural networks on HPC systems;and on adam trained models and a parallel method to improve the generalization performance.
Mental Stress has become a heightened concern in recent times. Stress creates physiological shifts that manifest within the realms of the human physiological system. Biosignals are considered realistic biomarkers for ...
详细信息
ISBN:
(纸本)9798350385939;9798350385922
Mental Stress has become a heightened concern in recent times. Stress creates physiological shifts that manifest within the realms of the human physiological system. Biosignals are considered realistic biomarkers for measuring an individual's emotional state. Among the various physiological signals considered in the study of Stress, a negative emotion, Electrodermal Activity (EDA), stands out as a promising BioSignal measuring the electrical properties of the skin, which is directly or indirectly related to emotional arousal. Six Time Domain Features are extracted further. Unsupervised machinelearning techniques such as K-means clustering are employed to label stressed EDA data into three Stress states: 'Low,' ' Moderate,' and 'High.' Six different classifiers are used to check the classification accuracy of the three stress levels. The Decision Tree achieved the highest precision rate, followed by 93% accuracy with random forest and Naive Bayes and Support Vector machine with 86% accuracy. Through the lens of EDA, this study delves into a better understanding of patterns of Stress, revealing its physiological underpinnings to contribute to a deeper insight into human well-being.
Techniques of machinelearning (ML) have recently been widely used in several applications, but not much for embedded systems, particularly those of safety-critical functionality, and are expected to help solve comple...
详细信息
ISBN:
(纸本)9798350381993;9798350382006
Techniques of machinelearning (ML) have recently been widely used in several applications, but not much for embedded systems, particularly those of safety-critical functionality, and are expected to help solve complex PHM and RUL which are difficult to be overcome with traditional approaches. This paper proposes a novel approach to encompass data-driven system-level health management, which comprises three main processes: an exhaustive offline training using machinelearning techniques to derive a model capable of predicting the system of interest within fixed performance criteria;followed by its deployment on the targeted embedded system for real-time health assessment and finally the ability to provide a multi-step ahead forecasting for the system prognostics. The ability to monitor the current state of the systems and predict their behaviour becomes essential. Therefore, techniques to establish fault tolerance and fault prediction are required. Condition monitoring (CM), prognostics and health management (PHM) and remaining useful life (RUL) are enablers for fault detection, fault progression and system degradation. This paper presents a ML-based methodology for the PHM of embedded systems, aiming at generalization and real-time performance.
作者：
Chaudhary, MeenakshiBanga, Puneet
Ambala Haryana Mullana133207 India Jmit
Department Of Computer Science & Engineering Radaur India
Department Of Computer Science & Engineering Ambala Haryana Mullana India
Cloud computing is the biggest buzz in the computer world these days. Cloud computing means different things to different people. In this paper, we have discussed the basic concepts of Cloud computing includes Cloud c...
详细信息
Smart transportation is one of the main research areas making use of artificial intelligence in order to build more convenient and smart modern cities. Current direction and navigation applications have shown the succ...
详细信息
ISBN:
(纸本)9798350349603;9798350349597
Smart transportation is one of the main research areas making use of artificial intelligence in order to build more convenient and smart modern cities. Current direction and navigation applications have shown the successful utility of using machinelearning for traffic forecasting and direction computing. More safety-related applications are also in research, such as accident detection and road anomaly detection. In this paper, accident detection and provision in smart transportation based on machinelearning is presented. Five machinelearning algorithms are trained and evaluated for predicting accident severity. An accident risk detection navigation application is embedded with a machinelearning algorithm for providing better user navigation.
This paper presents a novel technique for diagnosing human bone fractures by integrating machinelearning with microwave near-field MIMO probes. The proposed technique leverages a UWB MIMO microwave probe to different...
详细信息
ISBN:
(纸本)9798350385939;9798350385922
This paper presents a novel technique for diagnosing human bone fractures by integrating machinelearning with microwave near-field MIMO probes. The proposed technique leverages a UWB MIMO microwave probe to differentiate between healthy and fractured bones. By estimating the differences in the reflection coefficient of the probe response for both normal and abnormal cases, the technique effectively distinguishes between healthy and non-healthy bones. To enhance the variation in the probe response, the S-parameters are taken with varying distances from the phantom model with values starting from 1mm. Several supervised machinelearning algorithms are applied where the three best-performing algorithms are selected, and further analysis is conducted to determine the most effective algorithm among them. Experimental results demonstrate the effectiveness and accuracy of the proposed technique in diagnosing human bone fractures, showcasing its potential for practical application in the medical field. Further the study explores the development of machinelearning models trained on microwave data to accurately identify and classify various fracture types, offering a non-invasive and potentially more accessible method for early detection.
Deep learning is a cutting-edge methodology that has been extensively applied in real-world applications to solve computer vision tasks. Nonetheless, the inherent challenges of deep learning models lie in their black-...
详细信息
ISBN:
(纸本)9798331518783;9798331518776
Deep learning is a cutting-edge methodology that has been extensively applied in real-world applications to solve computer vision tasks. Nonetheless, the inherent challenges of deep learning models lie in their black-box nature, rendering them opaque and hard to interpret. Recently, attention-based vision transformers have been introduced to overcome the black-box behaviour of deep learning models. Despite these advances, the decision-making process of the vision transformer is still challenging to interpret. learning classifier systems is a state-of-the-art rule-based evolutionary machinelearning technique that stands out for its ability to provide interpretable decisions. These systems generate niche-based solutions, require less memory, and can be trained using small data sets. We hypothesize integrating attention mechanisms into learning classifier systems, aiming to identify critical components in problem instances, link features to create simple patterns, and model hierarchical relationships in the data. The experimental results for binary-class image classification (cat and dog) tasks demonstrate that the novel system successfully ignores the irrelevant parts and pays attention to the salient features of cats and dogs. Crucially, the proposed system exhibits comparable performance accuracy to that of the state-of-the-art learning classifier systems.
暂无评论