ML tasks

Although every ML problem is more or less an optimization problem, the way they are solved can vary. In fact, learning tasks can be categorized into three types: supervised learning, unsupervised learning, and reinforcement learning.

Supervised learning

Supervised learning is the simplest and most well-known automatic learning task. It is based on a number of predefined examples, in which the category to which each of the inputs should belong is already known, as shown in the following diagram:

The preceding diagram shows a typical workflow of supervised learning. An actor (for example, a data scientist or data engineer) performs Extraction Transformation Load (ETL) and the necessary feature engineering (including feature extraction, selection, and so on) to get the appropriate data with features and labels so that they can be fed in to the model. Then he would split the data into training, development, and test sets. The training set is used to train an ML model, the validation set is used to validate the training against the overfitting problem and regularization, and then the actor would evaluate the model's performance on the test set (that is, unseen data).

However, if the performance is not satisfactory, he can perform additional tuning to get the best model based on hyperparameter optimization. Finally, he would deploy the best model in a production-ready environment. The following diagram summarizes these steps in a nutshell:

In the overall life cycle, there might be many actors involved (for example, a data engineer, data scientist, or an ML engineer) to perform each step independently or collaboratively. The supervised learning context includes classification and regression tasks; classification is used to predict which class a data point is a part of (discrete value). It is also used for predicting the label of the class attribute. On the other hand, regression is used for predicting continuous values and making a numeric prediction of the class attribute.

In the context of supervised learning, the learning process required for the input dataset is split randomly into three sets, for example, 60% for the training set, 10% for the validation set, and the remaining 30% for the testing set.

Unsupervised learning

How would you summarize and group a dataset if the labels were not given? Probably, you'll try to answer this question by finding the underlying structure of a dataset and measuring the statistical properties such as frequency distribution, mean, standard deviation, and so on. If the question is how would you effectively represent data in a compressed format? You'll probably reply saying that you'll use some software for doing the compression, although you might have no idea how that software would do it. The following diagram shows the typical workflow of an unsupervised learning task:

These are exactly two of the main goals of unsupervised learning, which is largely a data-driven process. We call this type of learning unsupervised because you will have to deal with unlabeled data. The following quote comes from Yann LeCun, director of AI research (source: Predictive Learning, NIPS 2016, Yann LeCun, Facebook Research):

"Most of human and animal learning is unsupervised learning. If intelligence was a cake, unsupervised learning would be the cake, supervised learning would be the icing on the cake, and reinforcement learning would be the cherry on the cake. We know how to make the icing and the cherry, but we don't know how to make the cake. We need to solve the unsupervised learning problem before we can even think of getting to true AI".

The two most widely used unsupervised learning tasks include the following:

• Clustering: Grouping data points based on similarity (or statistical properties). For example, a company such as Airbnb often groups its apartments and houses into neighborhoods so that customers can navigate the listed ones more easily.
• Dimensionality reduction: Compressing the data with the structure and statistical properties preserved as much as possible. For example, often the number of dimensions of the dataset needs to be reduced for the modeling and visualization.
• Anomaly detection: Useful in several applications such as identification of credit card fraud detection, identifying faulty pieces of hardware in an industrial engineering process, and identifying outliers in large-scale datasets.
• Association rule mining: Often used in market basket analysis, for example, asking which items are brought together and frequently.

Reinforcement learning

Reinforcement learning is an artificial intelligence approach that focuses on the learning of the system through its interactions with the environment. In reinforcement learning, the system's parameters are adapted based on the feedback obtained from the environment, which in turn provides feedback on the decisions made by the system. The following diagram shows a person making decisions in order to arrive at their destination. Let's take an example of the route you take from home to work:

In this case, you take the same route to work every day. However, out of the blue, one day you get curious and decide to try a different route with a view to finding the shortest path. Similarly, based on your experience and the time taken with the different route, you'd decide whether you should take a specific route more often. We can take a look at one more example in terms of a system modeling a chess player. In order to improve its performance, the system utilizes the result of its previous moves; such a system is said to be a system learning with reinforcement.

So far, we have learned the basic working principles of ML and different learning tasks. However, a summarized view of each learning task with some example use cases is a mandate, which we will see in the next subsection.

Summarizing learning types with applications

We have seen the basic working principles of ML algorithms. Then we have seen what the basic ML tasks are and how they formulate domain-specific problems. However, each of these learning tasks can be solved using different algorithms. The following diagram provides a glimpse into this:

Types of learning and related problems

The following diagram summarizes the previously mentioned ML tasks and some applications:

ML tasks and some use cases from different application domains

However, the preceding diagram lists only a few use cases and applications using different ML tasks. In practice, ML is used in numerous use cases and applications. We will try to cover a few of those throughout this book.