ChatGPT is an Artificial Intelligence (AI) chatbot platform developed by OpenAI. Several studies have highlighted the advantages and disadvantages of integrating ChatGPT into teaching methodologies and knowledge gener...
详细信息
This paper analyses an electrification strategy for a depot for a public transport company, through an optimized model. The case study analyses both the transition of a depot able to host 10 buses towards e-buses and ...
详细信息
Post-training quantization of Large Language Models (LLMs) is *** this work, we introduce Low-rank Quantization Error Reduction (LQER), which combines quantization and low-rank approximation to recover the model *** l...
Post-training quantization of Large Language Models (LLMs) is *** this work, we introduce Low-rank Quantization Error Reduction (LQER), which combines quantization and low-rank approximation to recover the model *** leverages an activation-induced scale matrix to drive the singular value distribution of quantization error towards a desirable distribution, which enables near-lossless W4A8 quantization on various LLMs and downstream tasks without the need for knowledge distillation, grid search, or gradient-based iterative *** existing methods, the computation pattern of LQER eliminates the need for specialized Scatter and Gather processes to collect high-precision weights from irregular memory *** W4A8 LLMs achieve near-lossless performance on six popular downstream tasks, while using 1.36× fewer hardware resources than the leading state-of-the-art *** open-sourced our framework at ***/ChengZhang-98/lqer. Copyright 2024 by the author(s)
Face mask wearing detection technology can effectively supervise and control people wearing masks. The traditional mask detection uses only binary classification labels for training, which has low precision and recall...
详细信息
With the the vigorous development of new power system, energy and carbon coupling trading has emerged as an effective way to promote the renewable generation and reduce carbon footprint. It is a challenge to design an...
详细信息
Global navigation satellite system-reflectometry (GNSS-R), as an emerging observation method, has recently been applied to the retrieval of swell height. Existing research typically uses feature observables extracted ...
详细信息
The United States Department of Defense (DoD) has long been interested in aerial swarms for military intelligence, surveillance, and reconnaissance (ISR) missions. Through work with DoD partners, we found a unique nee...
详细信息
Physical layer secret key generation (PSKG) leverages the reciprocal properties of wireless channels to establish cryptographic keys between legitimate users. However, in time-division duplex (TDD) systems, this recip...
详细信息
UML use cases are commonly used in software engineering to specify the functional requirements of a system since they are an effective tool for interacting with stakeholders thanks to the use of natural languages. How...
详细信息
In an era driven by technological evolution, this paper embarks on an unprecedented journey to revolutionize rail track anomaly detection. Amidst the bustling world of transportation, our innovative approach harnesses...
详细信息
ISBN:
(数字)9789819718412
ISBN:
(纸本)9789819718405
In an era driven by technological evolution, this paper embarks on an unprecedented journey to revolutionize rail track anomaly detection. Amidst the bustling world of transportation, our innovative approach harnesses the synergy of cloud-based processing, audio analysis, and Mel-frequency cepstral coefficients (MFCC) extraction to unveil the previously hidden secrets of rail track conditions. The hallmark of our method lies in its dynamic fusion of intricate components. It transcends traditional boundaries and converts audio data into spectrograms through the short-time Fourier transform (STFT). This visual tapestry of frequencies unfolds a tale of spectral evolution across time, acting as the precursor to the essence of rail track anomalies. The proposed method commences with the conversion of audio data into spectrograms using short-time Fourier transform (STFT), facilitating the visualization of frequency content changes over time. Through this, the study computes MFCC features by first calculating Mel frequencies and subsequently deriving coefficients using cosine functions. Embodying key spectral characteristics, these features are then stored and standardized within a data frame. Enter cloud computing—a celestial realm of limitless computational prowess. Our approach transcends conventional confines by fusing the cloud's scalable might with the agility of audio analysis. Massive datasets unravel effortlessly, invoking a novel era of real-time rail track surveillance. Machine learning models, nurtured by standardized MFCCs, become vigilant sentinels against anomalies, forging a vigilant shield of safety. With experimental applause, our approach emerges victorious. A parade of metrics—accuracy, precision, recall, the cadence of F-score, and the majestic ROC curve—testify to the unparalleled prowess of our method. Anomalies that once whispered are now boisterously identified, etching a new chapter in rail track safety. In summary, this paper pioneers an innovati
暂无评论