The tourism industry is experiencing rapid growth and is becoming one of the fastest-expanding sectors. A considerable number of travelers now make hotel bookings and share their experiences on travel e-commerce platf...
详细信息
ISBN:
(数字)9798350352412
ISBN:
(纸本)9798350352429
The tourism industry is experiencing rapid growth and is becoming one of the fastest-expanding sectors. A considerable number of travelers now make hotel bookings and share their experiences on travel e-commerce platforms. Enhancing the quality of products and services within this industry can be accomplished by scrutinizing customer feedback. This research employs advanced deep learning techniques to discern subtle sentiments and glean insights from an extensive collection of hotel reviews. Deep neural networks, such as Convolutional Neural networks (CNNs) and Long Short-Term Memory (LSTM), are widely acknowledged for their effectiveness in sentiment analysis. This study conducts a comparative examination of various networkarchitectures, encompassing CNN, LSTM, CNN-LSTM, BiLSTM, and ConvBiLSTM, with the objective of pinpointing the most suitable approach. The experimental results are derived from datasets containing hotel reviews from Indonesia, obtained by crawling TripAdvisor. Notably, the LSTM model achieved an accuracy score of 96.42% on the Padma Hotel dataset and 85.31% on the Hard Rock Hotel dataset. The CNN-LSTM model demonstrated an accuracy of 85.87% on the Ayana Hotel dataset, while the BiLSTM model achieved 86.09% accuracy on the Pullman Hotel dataset. In assessing the performance of machine learning versus deep learning models, the analysis extends to commonly used IMDb datasets. The experimental results underscore the superiority of deep learning models over machine learning models across all evaluated metrics.
The penetration of embedded generation, including renewable power sources (wind and solar), is gradually increasing in power distribution networks. Also, the transition from conventional fossil fuel-based transportati...
The penetration of embedded generation, including renewable power sources (wind and solar), is gradually increasing in power distribution networks. Also, the transition from conventional fossil fuel-based transportation to e-transportation has introduced electric vehicle charging stations as a new load class. The conventional distribution system architecture alteration has made the system operation rather challenging. Therefore, an efficient energy management scheme is crucial to the satisfactory operation of an active distribution system from techno-economic considerations. This paper proposes an optimal operating strategy to simultaneously minimize the operating cost, average voltage deviation, and line loadings and improve the voltage stability of an active distribution network. The distribution system is assumed to have a soft open point and smart transformer for smooth active and reactive power control. The demand response flexibility (offered by responsive electrical demands and public and residential electric vehicle charging stations) is coordinated by controlling a smart transformer and a soft open point to realize multiple objectives. The multi-objective problem is solved in the fuzzy domain using a combination of linear programming and particle swarm optimization. Simulation results on a sixty-nine-bus radial distribution system validate the proposed method's effectiveness.
OPC UA is an open set of specifications for communication and modelling that have found extensive use in industrial automation. OPC UA information models are part of the specification and are used by asset owners and ...
详细信息
ISBN:
(纸本)9781728129891
OPC UA is an open set of specifications for communication and modelling that have found extensive use in industrial automation. OPC UA information models are part of the specification and are used by asset owners and equipment vendors to describe the static parameters and time varying signals belonging to equipment in standardized ways. Such descriptions are aimed to drastically reduce the cost of integrating software managing these industrial assets. Hindering the adoption of OPC UA information models is their inherent complexity and the problematic specification of query support. Previous research has implemented the OPC UA query specification using SPARQL and found that formulating queries in SPARQL natively provides shorter queries that are easier to formulate. However, parts of the OPC UA graph are time-varying, and may index large time series data sets, which are sub-optimally handled by most available SPARQL databases. We show how OPC UA can be mapped to the Semantic Web in a way that allows a hybrid approach where time series data is stored in a specialized system and describe a query rewriting approach that allows for joint queries over static OPC UA information models and time series data. Compared to existing approaches, we argue that our approach is better suited for analytics applications that require portability across infrastructure and allows a greater range of SPARQL engines. A prototype implementation called Quarry is open sourced under the Apache 2.0 license to improve the accessibility and reproducibility of this research area.
Feature fusion, the combination of features from different layers or branches, is an omnipresent part of modern networkarchitectures. It is often implemented via simple operations, such as summation or concatenation,...
详细信息
ISBN:
(纸本)9780738142661
Feature fusion, the combination of features from different layers or branches, is an omnipresent part of modern networkarchitectures. It is often implemented via simple operations, such as summation or concatenation, but this might not be the best choice. In this work, we propose a uniform and general scheme, namely attentional feature fusion, which is applicable for most common scenarios, including feature fusion induced by short and long skip connections as well as within Inception layers. To better fuse features of inconsistent semantics and scales, we propose a multi-scale channel attention module, which addresses issues that arise when fusing features given at different scales. We also demonstrate that the initial integration of feature maps can become a bottleneck and that this issue can be alleviated by adding another level of attention, which we refer to as iterative attentional feature fusion. With fewer layers or parameters, our models outperform state-of-the-art networks on both CIFAR-100 and ImageNet datasets, which suggests that more sophisticated attention mechanisms for feature fusion hold great potential to consistently yield better results compared to their direct counterparts. Our codes and trained models are available online(1).
Deep Learning and Machine Learning algorithms achieved good results in various tasks, from computer vision and speech recognition to condition monitoring and predictive maintenance. Besides the good results obtained, ...
详细信息
ISBN:
(数字)9798350380903
ISBN:
(纸本)9798350380910
Deep Learning and Machine Learning algorithms achieved good results in various tasks, from computer vision and speech recognition to condition monitoring and predictive maintenance. Besides the good results obtained, most Deep Learning algorithms do not provide a precise interpretation of their decisions. For the purpose of interpretability, explainable ML has gained attention. Because hardware accelerators do not support most explainable Machine Learning algorithms, they frequently experience runtime deficiencies. The resulting need for comparably inefficient software implementations limits their use on edge-computing hardware. This paper presents a method to represent trained explainable ML algorithms as Deep Neural networks to utilize hardware acceleration to meet the growing demand for accelerated inference of Machine Learning algorithms on edge hardware. The primary approach of the paper is to disassemble the trained explainable ML algorithms' inference into their basic mathematical operations to represent them as Deep Neural network layers. The techniques to convert the trained model functionalities to Deep Neural network layers are described in detail, including the layer functionalities and their usage in the Deep Neural networks. Due to the wide use of Deep Neural networks on hardware accelerators, this method allows the usage of affordable and efficient edge hardware instead of high-price, customized hardware and replaces the programming of high-effort compiler. Finally, the method is successfully applied to a part of an open-source ML toolbox, and the resulting Deep Neural network inferences are successfully run on a Neural Processing Unit.
Optimal control of stochastic systems involves finding control strategies that optimize certain performance criteria while accounting for the parametric uncertainties and stochastic additive disturbances involved in t...
详细信息
ISBN:
(数字)9781665410205
ISBN:
(纸本)9781665410212
Optimal control of stochastic systems involves finding control strategies that optimize certain performance criteria while accounting for the parametric uncertainties and stochastic additive disturbances involved in the system dynamics. Model predictive control (MPC) solves an open-loop constrained stochastic optimal control problem repeatedly in a receding-horizon manner, resulting in large computation sometimes. Alternatively, the proposed stochastic self-optimizing control (SOC) selects optimal nonlinear controlled variables (CVs) offline by minimizing the expectation of the weighted closed-loop loss function based on neural network training. The nonlinear self-optimizing CVs are simply kept constant online so that the satisfactory control performance can be achieved. The proposed stochastic SOC requires much less online computation time compared with MPC, which is demonstrated by a two-mass spring simulation model.
The evolving network infrastructure, particularly the 5G core network, is increasingly adopting cloud technologies. This shift brings to the forefront the challenge of meeting the demanding per-packet processing requi...
详细信息
ISBN:
(数字)9798350386059
ISBN:
(纸本)9798350386066
The evolving network infrastructure, particularly the 5G core network, is increasingly adopting cloud technologies. This shift brings to the forefront the challenge of meeting the demanding per-packet processing requirements posed by multi-hundred Gbps Ethernet NICs (network interface cards). While traditional NFV (network function virtualization) platforms are effective on older hardware, the per-packet run-to-completion (RTC) execution model for per-packet processing suffers from stalling on state access due to L1/L2 cache misses. Although previous work applying software prefetching can mitigate the issues, their applications are fundamentally limited by the nature of a single execution stream, hence limiting them to batch lookups, suffering from control-flow divergence, and requiring manual tuning. To address the limitations, we introduce a novel interleaved function stream execution model that exploits the function-level parallelism through memory-level parallelism, targeting feature-rich network functions such as 5G Core. To provide the visibility into network functions, we introduce a novel programming model based on the principle of Granular Decomposition, which provides deep visibility into the state access by decoupling the state in a more fine-grained manner compared to traditional modular approaches. We integrate these two innovative designs into a new open-source NF platform, which we refer to as GuNFu. We have tested GuNFu on widely deployed network functions such as 5G UPF (User Plane Function), 5G AMF (Access Management Function), NAT (network Address Translator) and others. Extensive evaluations reveal that GuNFu can achieve throughput ranging from 1.5 to 6 times over the traditional modular approach.
Applications of Big Data to smart cities are nearly limitless. However, challenges emerged with handling data with new magnitude of dimensionality, heterogeneity, required processing timeliness, and the lack of report...
详细信息
ISBN:
(纸本)9781665400589
Applications of Big Data to smart cities are nearly limitless. However, challenges emerged with handling data with new magnitude of dimensionality, heterogeneity, required processing timeliness, and the lack of reported development experiences on how to turn Big Data into Smart Data may hinder the adoption of Big Data technologies in delivering smarter city services. This paper reports our experience in developing a Smart Big Data-centric software platform for tracking city performance. Here we examine how to design such a platform by exploiting architectural styles and open source technology for Big data management. We describe the actual development of an instance of it, the PELL Smart City Platform, for processing and managing urban data in the domain of public lighting. In particular, we focus on the formulation and evaluation of key performance indicators related to energy consumption to derive Smart Data from Big Data in the context of public street lighting.
Many waveform co-design studies demonstrate theoretical performance enhancements but rarely provide a clear path toward implementing a tractable solution. To address this limitation, we developed a waveform co-design ...
Many waveform co-design studies demonstrate theoretical performance enhancements but rarely provide a clear path toward implementing a tractable solution. To address this limitation, we developed a waveform co-design technique to maximize the joint radar-communications network's joint per-formance and a computationally tractable method for optimizing it. This waveform co-design technique is based on the theory of partially-observable Markov decision processes (POMDPs), which we solve using an approximate dynamic programming approach called nominal belief-state optimization (NBO). The POMDP framework's natural look-ahead feature allows us to trade between the short-term and long-term performance of both radar and communications tasks, which allows it to adapt to changes in system requirements and environmental conditions. Using the WISCANet over-the-air experimental radio testbed, we implement a simple joint radar-communications system and demonstrate this waveform design and optimization technique in a hardware-in-the-loop over-the-air demonstration. We further extend the problem by proposing a real-time waveform optimization solution using a Kalman Filter in a dynamic environment.
Citation recommendation focuses on recommending references to the given document automatically. While the existing works have achieved improvements in citation recommendation, their quality may still be degraded due t...
详细信息
Citation recommendation focuses on recommending references to the given document automatically. While the existing works have achieved improvements in citation recommendation, their quality may still be degraded due to the cold-start problem or insufficient use of features. To alleviate these problems, we investigate the possibility of fully utilizing a document's features from different dimensions. Specifically, we present a citation recommendation model based on Siamese BERT and Knowledge Graph (CRM-SBKG), a mixture model that mines the text feature and entity feature to obtain the recommendation list. We use the BERT model with the Siamese network for text feature extraction and build a citation knowledge graph where entity feature is extracted. We apply the proposed model on open Academic Graph (OAG) and DataBase systems and Logic programming (DBLP). The results show that our model outperforms the baselines significantly. It reflects that the proposed model can extract features more effectively and make better citation recommendations.
暂无评论