Introduction to Artificial Intelligence and Machine Learning

Original article was published by manasvi logani on Artificial Intelligence on Medium


Some people call this artificial intelligence, but the reality is this technology will enhance us. So instead of artificial intelligence, I think we’ll augment our intelligence.” — Ginni Rometty

What is Artificial Intelligence?

Every species survive on the basis of the type of intelligence they possess. Homo Sapiens are termed as the most “intelligent” species of all because their intelligence consists of many additional and more advanced qualities as compared to others.

These qualities are:

  1. Estimation or PredictionIt is the quality of forecasting a result on the basis of the factors affecting them.
  2. Differentiation or Classification- It is the quality of categorizing different items into definite classes, for example, cats and dogs.
  3. Segmentation- It is the quality of grouping similar things together, for example, plants and animals.
  4. Recommendation- It is the quality of recommending something to someone according to their interests.
  5. Decision Making- It is the quality of making a choice on the basis of the information gathered.

Now, when a machine is made or taught to mimic the human tendency to fulfill the above five qualities of intelligence, it is termed as Artificial Intelligence.

What is Machine Learning?

Machine learning is a subset of Artificial Intelligence. In simple words, it consists of Algorithms or models which helps us achieve Artificial Intelligence.

It is an intersection of the field of Mathematics, Computer Science and Optimization.

Types of machine learning:

Supervised Machine Learning

Image Source: freepik.com

Supervised learning, as the name suggests is derived from the word supervisor. It is a person who supervises on another person or an activity. A teacher is an example of a supervisor provides guidance.

For example, when we were taught addition, we were given a method and some examples. We learnt the process using this and were given other data similar to the provided examples to apply our knowledge of addition.

Similarly, in supervised learning we teach the machine using some examples called a labelled training dataset, through a method or an algorithm. Now, after the machine learns using this; unknown data called the test dataset is introduced for the machine to provide an unbiased evaluation of a final model fit on the training dataset.

Examples: Regression(algorithms like Linear Regression, Polynomial Regression), Classification(algorithms like Support Vector Machines, Random Forest)

Applications: Speech Recognition, Spam detection etc.

Unsupervised Machine Learning

Image Credits: TechNative.com

In Unsupervised Learning, no guiding force is present to train the machine. The data provided is unstructured and unlabeled on which the algorithm is allowed to act upon. The machine has to identify patterns in the data by itself. The main goal is to find the similarities and differences between data points.

For example, a dataset contains unlabeled image dataset of three fruits- apple, banana and strawberry. The machine cannot identify the type of fruit by looking at the picture. Without any labels guiding it, using an algorithm, the machine has to identify different features, hence categorizing them into clusters.

Examples: Clustering (algorithms like KNN, K- means clustering etc.)

Applications: Image processing, pattern recognition etc.

Reinforcement Learning

Image Credits: wikipedia.com

Reinforcement Learning is driven by the need to maximize reward by taking suitable actions. In this type, there is no training dataset with clear answers, instead the machine works through gaining experience.

For example, a robot has been introduced to an environment with gold as reward and thorns as negative agents which it has to dodge. If the robot comes in contact with the thorns, then some points will be deducted but if it takes the correct path to reach the final gold reward, some points will be rewarded. The final goal is to achieve the maximum reward possible.

Example : Q- learning

Applications: Games, Robotics, Aircraft control etc.