*3.1. Machine Learning Types*

#### 3.1.1. Supervised Learning

In this ML type, models are defined in advance and systems learn from the given input and output pairs, i.e., the input data and desired output are labeled [61]. With enough data knowledge, one can help the machine connect the dots with supervised learning using the labeled sample data and correct output.

### 3.1.2. Unsupervised Learning

Here, the AI learns without the aid of predefined target values, i.e., the model is required to identify patterns in an unlabeled input data [61]. Learning and improving by trial and error is key to unsupervised learning. Unlike supervised learning, here you are not working with labeled data, you are not showing the machine the correct output. You are using different algorithms to let the machine connect the dots by studying and observing data. In unsupervised learning, the chances of the machine to find patterns or classifications that humans can never see is very high.

#### 3.1.3. Semi-Supervised Learning

This is a combination of supervised and unsupervised learning advantages [61]. Here, training starts with a small amount of dataset in order to allow the machine to get familiarized with the data. In addition, the machine studies and observes the data to expand its vocabulary/database using inductive reasoning. Another form of semisupervised learning is the transductive reasoning, which allows one to narrow down the unlabeled data using unknown knowledge of collected data. Semi-supervised learning is not very common in machine learning applications.

#### 3.1.4. Reinforcement Learning

In this situation, the model is granted autonomy to engage with a dynamic environment that gives feedback based on rewards and punishments, i.e., the model is taught through positive and negative interactions [61]. This method of learning differs significantly from the other three methods. The machine iterates until the outcome is enhanced each time, coming closer and closer to high-quality output.

#### 3.1.5. Multitask Learning

Multitask learning helps several algorithms share their experience with each other, thereby helping them learn concurrently rather than individually [73].

#### 3.1.6. Ensemble Learning

Ensemble learning is a combination of two or more algorithms that form one single algorithm [74]. Here, it has been observed that a collection of algorithms almost always outperforms an individual algorithm when carrying out a particular task [75].
