We are living in the age of Artificial Intelligence and Machine Learning. AI and ML are probably the two most popular tech buzzwords of the current time. But, what do the terms AI, ML, (and, Deep Learning) actually mean? Even though we often hear people uttering these three term in the same breath, do they mean one and the same thing? Or, are there any differences between them, however subtle those might be? This technical article attempts to answer some of such common questions that revolve around in people's minds more often than not.

If we were to take a formal definition of Machine Learning (ML), it would be - Machine learning is the subfield of computer science that gives "computers the ability to learn without being explicitly programmed”. Machine Learning algorithms have been inspired by nothing other than the human learning process - the algorithms iteratively learn from the data and allow computers to find hidden insights. In the field of Machine Learning, we build something called 'Machine Learning Models' - the models are built using Machine Learning Algorithms and they help us in a variety of tasks such as Object Recognition, Recommendation, Classification, to name a few.

There are many different applications of Machine Learning that we see everyday in our daily lives - chatbots, computer games using face recognition, Netflix and Amazon recommending videos, movies and TV shows to its users, and many more other applications. Each of these applications use different machine learning techniques and algorithms. A few of the more popular techniques include Regression/Estimation technique, Classification technique, Clustering technique, Association technique, Anomaly Detection technique, Sequence Mining technique, Dimensionality Reduction technique, Recommendation Systems, etc.

1. Regression/Estimation technique: This technique is used for predicting a continuous value - for example, predicting things like the price of a house based on certain characteristics like the size of the house and its location, or to estimate the CO2 emission from a car's engine.
2. Classification technique: A Classification technique is used for predicting the class or category of a case, for example, if a cell is benign or malignant, or whether or not a customer will churn.
3. Clustering technique: This technique is used for clustering groups of similar cases, for example, finding similar patients, or customer segmentation in the banking field.
4. Association technique: Association technique is used for finding items or events that often co-occur, for example, grocery items that are usually bought together by a particular customer.
5. Anomaly Detection technique: Anomaly detection is used to discover abnormal and unusual cases, for example, credit card fraud detection.
6. Sequence Mining technique: Sequence mining is used for predicting the next event, for instance, the click-stream in websites.
7. Dimensionality Reduction technique: Dimensionality reduction is used to reduce the size of data.
8. Recommendation Systems: Recommendation systems associate people's preferences with others who have similar tastes, and recommends new items to them, such as books or movies.

At this point, I want to answer a question that has probably crossed our mind at some point or the other. These days, we keep on hearing the buzzwords 'Artificial Intelligence (AI)', 'Machine Learning (ML)' and 'Deep Learning'. Even if minute, but these three terms do have certain differences.

Artificial Intelligence (AI) tries to make computers intelligent in order to mimic the cognitive functions of humans. So, Artificial Intelligence is a general field with a broad scope including Computer Vision, Language Processing, Creativity, and Summarization.

Machine Learning (ML) is the branch of AI that covers the statistical part of artificial intelligence. It teaches the computer to solve problems by looking at hundreds or thousands of examples, learning from them, and then using that experience to solve the same problem in new situations.

And, Deep Learning is a very special field of Machine Learning where computers can actually learn and make intelligent decisions on their own. Deep learning involves a deeper level of automation in comparison to most of the machine learning algorithms.


I will next touch upon the concepts of supervised learning/algorithms versus unsupervised learning/algorithms very briefly. If you take a careful look at the words that make up the terms 'supervised learning/algorithms' and 'unsupervised learning/algorithms', you will probably get a fair amount of hint at what these terms refer to.

In plain English language, the word 'supervise' means to observe and direct the execution of a task, project, or activity. In the world of Artificial Intelligence and Machine Learning, we are not going to supervise a person obviously; as you might have already guessed by now, we will be supervising a machine learning model that might be able to perform a certain task intelligently for us, classifying images, for example. So, how do we supervise a machine learning model? We do this by “teaching” the model, that is, we load the model with knowledge so that we can have it predict future unknown instances. And, this leads to the next question, which is, “How exactly do we teach a machine learning model?” - we teach the model by training it with some data from a labeled dataset. Please make a note of the fact that it is really important that the data be labeled. Now, yet another question comes to our mind - what does a labeled dataset look like? Let's say that we have a dataset containing numerous records and each record represents historical medical data of patients. For each record, we know whether or not the patient was diagnosed with Cancer. Thus, we know the class of each record and that is what we call a labeled record. And, a labeled dataset is nothing but a collection of numerous labeled records.

Let us have a look at a sample labeled dataset.


Let’s get introduced to some of the components of the dataset. The names - Clump, UnifSize, UnifShape, MargAdh, and so on, are called ATTRIBUTES. The columns are called FEATURES and they include the data. A particular row in the dataset is referred to as an OBSERVATION.

Looking directly at the value of the data, we have two kinds. One of the types is numerical data. When dealing with machine learning, the most commonly used data is the numeric data type. The second data type is categorical; that is, it is non-numeric, because it contains characters rather than numbers. In the case of our sample dataset, it is categorical because the dataset is made for Classification - whether a particular patient has a benign or malignant tumour.


Now, there are two types of Supervised Learning techniques. They are: classification and regression. Classification is the process of predicting a discrete class label or category. Regression is the process of predicting a continuous value as opposed to predicting a categorical value in Classification. For example, if you have a dataset that contains data pertaining to the CO2 emissions from the various models of automobiles and includes attributes like engine size, number of cylinders, fuel consumption and CO2 emissions from different cars, we can use regression to predict the CO2 emission of a new car model by using other fields, such as engine size or number of cylinders.


Since we now know the meaning of supervised learning, it is pretty easy to guess what unsupervised learning means. Yes! Unsupervised Learning is exactly as it sounds - we do not supervise the machine learning model, but we let the model work on its own to discover information that may not be visible to the human eye. It means that the unsupervised learning algorithm trains on the dataset, and draws conclusions on UNLABELED data. Generally speaking, unsupervised learning has more difficult algorithms than supervised learning, since we know very little to no information about the data, or the outcomes that are to be expected. Dimensionality reduction, Density estimation, Market basket analysis and Clustering are some of the most widely used unsupervised machine learning techniques.

• Dimensionality Reduction and/or feature selection helps reduce redundant/unimportant features to make the task of classification easier.
• Market basket analysis is a modelling technique based upon the theory that if you buy a certain group of items, you are more likely to buy another group of similar/related items.
• Density estimation is a very simple concept that is mostly used to explore the data to find some form of structure within it.
• And finally, clustering. Clustering is considered to be one of the most popular unsupervised machine learning techniques used for grouping data points or objects that are somehow similar. Cluster analysis has many applications in different domains, whether it be a bank’s desire to segment its customers based on certain characteristics, or helping an individual organize and group his/her favourite types of music! Generally speaking, though, Clustering is used mostly for: Discovering structure, Summarization, and Anomaly detection.

So, before concluding, I will recap the important differences between Supervised and Unsupervised Learning. Supervised learning deals with labeled data while Unsupervised Learning deals with unlabeled data. In supervised learning, we have machine learning algorithms for Classification and Regression. In unsupervised learning, we have methods such as clustering. In comparison to supervised learning, unsupervised learning has fewer models and fewer evaluation methods that can be used to ensure that the outcome of the model is accurate. As such, unsupervised learning creates a less controllable environment, as the machine is creating outcomes for us.


The author of this technical article is an IBM Z Champion for 2020.