China
Africa
Explore chapters and articles related to this topic
New Approaches to Mobile Ad Hoc network Routing: Application of Intelligent optimization Techniques to Multicriteria Routing
Published in Jonathan Loo, Jaime Lloret Mauri, Jesús Hamilton Ortiz, Mobile Ad Hoc Networks, 2016
Bego Blanco Jauregui, Fidel Liberal Malaina
A neural network comprises a network of simple processing elements (the artificial neurons) that, as a group, produce a complex global behavior thanks to a mathematical or computational model that makes it possible to process the information received by the system. Often, a neural network is an adaptive system that modifies its structure as the internal or external information flows through the network. The final result is that the neural network is able to infer a function from the observation and then use it to generate new results. In summary, it learns from the observation [79]; i.e., given a task that must be worked out and a class of functions F, the neural network employs observation to find the function f ∈ F that optimally solves the task.
IoT Recommender System: A Recommender System Based on Sensors from the Internet of Things for Points of Interest
Published in Vijender Kumar Solanki, Vicente García Díaz, J. Paulo Davim, Handbook of IoT and Big Data, 2019
Cristian González García, Daniel Meana-Llorián, Vicente García Díaz, Edward Rolando Núñez-Valdez
To achieve its goal, AI usually relies on different approaches. For example, search and optimization algorithms are important for reasoning and leading from premises to conclusions. Fuzzy logic is a version of first-order logic which allows the truth of a statement to be represented as a value between 0 and 1, rather than just 1 or 0. Probabilistic methods (e.g., Bayesian networks, Hidden Markov model, etc.) are very interesting for uncertain reasoning. Classifiers and statistical learning methods such as machine learning give computers the ability to learn without being explicitly programmed. Neural networks are computational models used to solve problems in the same way that the human brain would do.
Conceiving Design Solutions
Published in Bahram Nassersharif, Engineering Capstone Design, 2022
If the design solution is for a process or a system that cannot be physically modeled, then a computational model will be the preferred solution. Computational models are input to software programs (computer codes) to create the geometry and associated boundary and initial conditions to simulate the performance of an existing or designed system for verification, validation, benchmarking, certification, parametric design, and insight.
On continuous health monitoring of bridges under serious environmental variability by an innovative multi-task unsupervised learning method
Published in Structure and Infrastructure Engineering, 2023
Alireza Entezami, Hassan Sarmadi, Bahareh Behkamal, Carlo De Michele
Machine learning is a key area of artificial intelligence that intends to develop computational models for learning from data (i.e. features) and perform accurate decision-making for some tasks such as classification, prediction, clustering, anomaly detection, etc. (Alpaydin, 2014). The great advantage of machine learning in SHM is to provide an automated strategy for safety and health assessment of any kinds of civil structures in long- and short-term monitoring programs. Machine learning-aided SHM is often divided into three main categories of supervised, semi-supervised, and unsupervised learning. In these categories, one attempts to learn a model by using training data and make a decision via test data. However, the application of these algorithms relies strongly on the availability of the labels of training samples.
Client profile prediction using convolutional neural networks for efficient recommendation systems in the context of smart factories
Published in Enterprise Information Systems, 2022
Nadia Nedjah, Victor Ribeiro Azevedo, Luiza De Macedo Mourelle
Machine learning uses predictive computational models and methods created from observational data. In this context, the machine develops the ability to continuously learn from data, managing predictions and pattern recognition just like humans do (Louridas and ebert 2016). Machine learning is applied to provide a smart solution for many existing computational problems, such as time series forecasting (Araújo et al. 2019; Araújo et al. 2018), computer vision (Pinheiro, Nedjah, and Mourelle 2020), document (Nedjah, Santos, and Mourelle 2020), among many others (Li et al. 2019; Sarivougioukas and Vagelatos 2020). From the efforts of corporations such as Google, Microsoft, Facebook and Amazon, machine learning has become one of the most important topics in computational science in the last decade (Edgar and Manz 2017).
NDE 4.0 compatible ultrasound inspection of butt-fused joints of medium-density polyethylene gas pipes, using chord-type transducers supported by customized deep learning models
Published in Research in Nondestructive Evaluation, 2020
Maryam Shafiei Alavijeh, Ryan Scott, Fedar Seviaryn, Roman Gr. Maev
Machine learning is a field of AI in which computational models are trained to conduct tasks automatically through experience with data [14]. Deep learning is a type of machine learning that specifically uses deep artificial neural networks [15,16] and has shown excellent, even superhuman performance in a variety of tasks [14] and is thus becoming increasingly prevalent in NDE applications [17–23]. Machine learning can be broadly categorized into supervised learning and unsupervised learning; the former requires both input and target output vectors to be provided to the model during training while the latter only requires input vectors without target outputs. Supervised learning can be further subdivided into classification and regression, while unsupervised learning typically involves clustering or representation learning. Flaw detection is often framed as classification or object detection when labeled data are cheap and easy to obtain for all the required classes [24–27]. However, deep learning models are notorious for requiring immense amounts of data. In supervised learning, the issue is exacerbated as data labeling can often be one of the most expensive and time-consuming processes in machine learning. In the context of flaw detection, it is often difficult to reliably simulate the various types of flaws that are observed in the production environment. Thus, in such situations, unsupervised deep learning approaches are often useful.