Green Internet of Things and Machine Learning. Группа авторов
Чтение книги онлайн.
Читать онлайн книгу Green Internet of Things and Machine Learning - Группа авторов страница 11
Название: Green Internet of Things and Machine Learning
Автор: Группа авторов
Издательство: John Wiley & Sons Limited
Жанр: Программы
isbn: 9781119793120
isbn:
1.2.2.4 Reinforcement Learning
Reinforcement learning does not need training examples. In the reinforcement learning, models are given an environment, group of some actions, a goal and a reward. This algorithm learns by rewards and penalties. For every correct output, a reward is given and a penalty for every wrong output. To produce the desired output, the algorithm has to maximize these rewards. It is named reinforcement learning because for every reward the model gets a reinforcement that it is on right path. The reward feedback system helps the model to predict future behavior [9]. Figure 1.4 shows the complete process of reinforcement learning.
The following are algorithms which are based reinforcement learning:
• State Action Reward State action (SARSA)
• Q-Learning
• Deep Q Neural Network (DQN)
1.3 Deep Learning
Deep Learning (DL) is the concept AI that acts like the human brain to process and creating the patterns which helps to take the decisions. It is a subset of ML in AI that has ability to learning from unsupervised, unlabeled or unstructured. DL is becoming more popular as it achieves high accuracy and helps us in making decisions, translating languages, detecting objects, and recognizing speech [10].
Figure 1.5 Correlation between AI, ML, and DL.
1.4 Correlation Between AI, ML, and DL
Figure 1.5 [9] depicts the correlation among ML, DL, and AI. Here, as we can see that DL is the subset of ML, and ML is the subset of AI. Hence, initially, AI came into the existence first, and later, ML erupted from it. To be more specific and denser, DL is derived from ML further.
1.5 Machine Learning–Based Smart Applications
1.5.1 Supervised Learning–Based Applications
1.5.1.1 Email Spam Filtering
It helps in filtering junk e-mail or unwanted commercial e-mail and bulk e-mail from the true e-mails. With the usage of these learning algorithms, spam filter helps the user not to be flooded with the bulk or junk e-mails. The spam filter learns by watching the pattern of genuine e-mails and junk e-mails [11].
1.5.1.2 Face Recognition
Human face is not unique. Various factors cause to vary the face. With the help of these learning algorithms, face recognition has become easier. Face recognition is used in various situations such as security measure at an ATM, criminal justice system, image tagging in social networking sites like Facebook, an image database investigation, and areas of surveillance [11].
1.5.1.3 Speech Recognition
To recognize the speech, the ML methods can be used. It involves two different learning phases: The first phase is speaker dependent where, after purchasing, the software user has to train the model by his/her voice to achieve accuracy, and in the second phase, before the software is shipped, the model is trained by default. It is speaker independent fashion [12].
1.5.1.4 Handwriting Recognition
Automated handwriting recognition through supervised ML really solves a complex problem of humans and cut down a large amount of time. Therefore, it is being utilized in various applications [12].
1.5.1.5 Intrusion Detection
Intrusion is the biggest problem of today’s era. When a person or a process wants to enter unauthorizedly into another network, it is known as Intrusion. Therefore, this intrusion detection is important to scrutinize and to identify the threats or violations to the computer security. Learning algorithms helps in finding the intrusion.
1.5.1.6 Data Center Optimization
Huge energy requirement and environmental responsibility are rising a pressure day by day to Data Center (DC) companies to keep a DC operating efficiently. The ML algorithms help the DC to monitor the energy consumptions and pollution levels relentlessly to improve the operating efficiency [13].
1.5.2 Unsupervised Learning–Based Applications
1.5.2.1 Social Network Analysis
Identification of a person with in a large or small circle on social media platforms such as Facebook and Instagram has become easier with the help of unsupervised learning. It assists in maintain the similar posts in the proper way [14].
1.5.2.2 Medical Records
Automation helped the medical industry to manage the records in better way. Now, e-medical records have turn out to be ubiquitous [15]. Therefore, medical data is getting shape of medical facts and surprisingly helping to understand the disease in better way.
1.5.2.3 Speech Activity Detection
Speech activity detection (SAD) helps to detect the presence or absence of human speech for speech processing. ML assists to reduce the unwanted noisy and long non-speech intervals from the speech. SAD helps in making human-computer interfaces. It helps the hearing-impaired people to use the machine or computer using the voice commands [16]. It is language independent program. SAD is having two types: supervised SAD and unsupervised SAD. Supervised SAD uses the available training data and models a system accordingly, while unsupervised SAD is a feature-based technique.
1.5.2.4 Analysis of Cancer Diagnosis
Nowadays, human life is being saved with the help of medical science and technology. Therefore, the contribution of technology to fight against the cancer is not surprising anymore. It is first step to find the type of cancer in order to cure it. Now, it is possible with the help of classification process by collecting patient samples. Some ML techniques like radial basis function (RBF), Bayesian networks, and neural networks trees are used to detect the cancer and its type [17].
1.5.3 Semi-Supervised Learning–Based Applications
1.5.3.1 Mobile Learning Environments
Mobile learning means with the help of mobile device and internet facility, we can learn anywhere any time. To learn from mobile, СКАЧАТЬ