Green Internet of Things and Machine Learning. Группа авторов
according to available datasets and implicitly comparing the current outcome to the final output [2].
1.2.1 Difference Between Artificial Intelligence and Machine Learning
In a general sense, AI and ML are much the same, but the fact is ML is the subset of AI as depicted in Table 1.1 [3].
1.2.2 Types of Machine Learning
• Supervised learning
• Unsupervised learning
• Semi-supervised learning
• Reinforcement learning
Figure 1.1 depicts the various types of machine learning techniques.
Table 1.1 Difference between AI and machine learning.
| Artificial Intelligence | Machine learning |
| AI enables the machines to behave or simulate like humans. | ML permits a machine to learn from available past data without giving instructions to it explicitly. |
| AI is used to make such systems which can solve complex problems like humans. | ML goal is to make a machine to be trained itself from historical data without any human intervention. |
| AI has ML and DL as subset. | ML has DL as subset. |
| Following three types of AI: general AI, strong AI, and weak AI. | Following four types of ML: semi-supervised, unsupervised, reinforcement, and Supervised learning. |
| AI focuses to maximize the chance of success. | Machine learning focuses on accuracy and patterns. |
| AI uses structured, unstructured data, and semi-structured. | ML uses structured and semistructured data only. |
1.2.2.1 Supervised Learning
In the supervised ML, a machine learns from past data and then produces the desired output [4]. A machine gets its training from already available dataset using appropriate algorithms and inferred function. This inferred function predicts the output and gives an approximate desired result. The used labeled data set helps the algorithm to understand the data and produce the labeled output for more accurate results [5]. Figure 1.2 shows the complete process of supervised learning.
The following are some algorithms which are based on supervised learning:
• Linear Regression
• Naive Bayes
• Nearest Neighbor
• Neural Networks
• Decision Trees
• Support Vector Machines (SVM)
Figure 1.1 Classification of machine learning.
Figure 1.2 Process of supervised learning.
1.2.2.2 Unsupervised Learning
When a machine learns from unlabeled data or it discovers the input pattern itself, it is known as unsupervised learning. It divides the learning data into diverse clusters. Therefore, this learning is known as clustering algorithm. In this learning, the training data will not be labeled and inferences functions create its own inferences by exploring the unlabeled dataset in order to find suitable patterns [6]. Figure 1.3 shows the complete process of unsupervised learning.
Name of common unsupervised algorithms:
• Anomaly detection
• K-means clustering
• Neural networks
• Hierarchal clustering
• Independent component analysis
• Principle component analysis
1.2.2.3 Semi-Supervised Learning
When the machine learns from both labeled and unlabeled data, it is known as semi-supervised learning. When it is not feasible to label the data due to lack of resource to label it or due to the large size of the data, semi-supervised learning is used [7]. It lies among the supervised and unsupervised learning. For the model building, semi-supervised learning is best. Semi-supervised learning makes use of small amount of labeled data but large amount of unlabeled data [8].
Figure 1.3 Process of unsupervised learning.
Figure 1.4 Process of reinforcement learning.
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