What is Machine Learning?

What is Machine Learning?

According to Tom Mitchell, a professor at Carnegie Mellon University has defined Machine Learning (ML) as, “A computer program is said to learn from experience E with some class of tasks T and performance measure P if its performance at tasks in T, as measured by P, improves with experience E.”

Consider playing checkers.
E = The experience of playing many games of checkers
T = The task of playing checkers.
P = The probability that the program will win the next game.

In 1959, Arthur Samuel – one of the pioneers of ML, defined machine learning as a “Field of study that gives computers the ability to learn without being explicitly programmed.” That is, ML programs have not been explicitly entered into a computer, like the if or then statements. ML programs, adjust themselves in response to the data they are exposed to. Samuel taught a computer program to play checkers better than himself which he couldn’t program it explicitly. However, in 1962 he succeeded, and his program beat the Connecticut state Checkers Champion.

However, ” ML is the science of making computers to learn and act like humans do, and enhance their learning in autonomously over time by providing them with data sets and information in the form of observations and real-world interactions.” ML, on the other hand, is solely focused on writing software that can learn from past its experiences. Machine learning is more closely related to data mining and statistics than it is in AI. Machine learning is a sub-field of AI. The goal of ML is to enable computers to learn on their own. A machine learning’s algorithm allows it to identify patterns in observed data, build models that explain the world, and predict things without having explicit pre-programmed rules and models.

The “learning” aspect of ML refers that, machine learning algorithms attempt of optimizing along a specific dimension; that is, they try to reduce the error and increase the likelihood of their predictions being right. The machine in ML refers to an algorithm that combines logic and math, or it is a method of computation. What it just means, if a computer program can improve how it performs a particular task based on its experience then we can say it has learnt. This is entirely different for a program which can perform a task just because its programmers have already defined all the parameters and data needed to perform that task. For, e.g., a computer program can play tic-tac-toe and crosses all the levels because a programmer wrote a code with a built-in winning strategy.

A program that has no predefined strategy and only a set of rules about the legal moves will require to learn by repeatedly playing the game until it can win. This doesn’t just apply to the games but is true of program performed classification and prediction classification. It is the process whereby a machine can recognize and categorize things from data sets including from the visual data and measurement data prediction known as a regression in statistics. Where a machine can guess and predict the value of something based on the previous values. For, e.g., Given a set of characteristics about a house, how much is it worth based on the last house sales and this leads us to another definition of ML – It is the extraction of knowledge from data. The most straightforward way of processing the Big data which is the self-adaptive algorithm that gets better and better patterns and analysis with every newly added data or with experience.

Machine Learning Algorithms can be divided into four main categories:

1. Supervised Machine Learning Algorithm (SMLA): In a supervised machine learning, the machine tries to learn from the previous examples of data sets that are fed into the system. In this method of machine learning, you teach, train the machine using data which is well labelled. In which the information is already tagged with the correct answer. SMLA can further be classified into two parts:
   • Classification: In classification, the problem is when an output variable is a group or a category, such as “white or black” or “spam and no spam.”
   • Regression: In a regression problem is when the output variable is a real-value such as “weight” or “dollars.”
2. Unsupervised Machine Learning Algorithms (UMLA): In unsupervised learning, the algorithms are left to discover themselves the new structures in the data-sets. Mathematically UMLA is when you only have an input data-set (A) with no corresponding output variables. In this learning algorithm, there are no given right answers, and the machine itself finds the answers. The UMLA can further be divided into two parts:
   • Association: The association rule learning is where you want to discover rules that describe the large portions of the data-sets, such as “People that tend to by A will also buy B.”
   • Clustering: The clustering rule is where you want to discover the natural groupings in the data-sets like a grouping of customers by their buying behavior.
3. Semi-Supervised Machine Learning Algorithms (SSMLA): The SSMLA comes somewhere in between the Supervised and Unsupervised machine learning as they use both the structured and unstructured data-sets for training which is typically a small amount of labelled and a large unlabeled data. The machines which use this method considerably improve their learning accuracy. This method of the learning algorithm is chosen when the structured data requires skilled and relevant experts to train them or learn from it.
4. Reinforcement Machine Learning Algorithms (RMLA): The computer program will interact with a dynamic environment in which it performs a particular task such as driving a car or playing a game with an opponent. The program is given feedback regarding a reward or punishment as it navigates its problem space. Using RMLA, the machine is trained to take specific decisions. The computer is exposed to an environment for continuously teaching itself using the trial and error methods.

Today, ML is being used in varied range applications. Facebook’s News Feed uses ML to personalize its every member’s feed. If a member stops frequently scrolling to read or like a particular friend’s post, the news feed will also start to show more of that friend’s activities in the feed. In this case, the software is using statistical analysis and predictive analytics in identifying patterns in the user’s data and use those patterns to populate the news feed. If the member continues to like, comment, and read on his friend’s posts, the new data will be included in the data sets and the news feed will adjust itself accordingly.

ML has entered an array of enterprise applications. Customer Relationship Management (CRM) systems use the learning models to analyse email and prompt the sales team in an organisation to respond to the critical messages first. Advanced systems can recommend effective responses too. Business Intelligence and analytics vendors are using ML in their software to help users to identify essential data points automatically. HR systems also use learning models for the identification of characteristics of efficient employees and apply this information to source the best applicants for an open position.

Similarly, ML plays a vital role in self-driving cars. Deep Learning Neural Networks are used to recognize objects and determine actions that are optimal for safely steering the vehicle on the road. Virtual Assistant technology is powered with ML. Smart Assistants is combined with several Deep Learning models for the interpretation of speech, voice recognition, highlighting the user’s schedule or preferences defined previously for taking action such as booking a flight or showing the driving directions.

To summarize, Machine Learning is a field which is a subset of Artificial Intelligence, and today the world is undergoing a massive technological revolution for which application of AI is critical for building smarter and intelligent machines which are going to disrupt the way the industry and this world will live, work, and perform their activities.

Image Credit: Pixabay