A Primer on AI
Artificial Intelligence (AI) is a domain of science and technology that tries to mimic natural intelligence using fundamentals of a plurality of natural phenomenon to solve a problem. The term artificial intelligence is made up of two words “Artificial” in context to “man-made” and “Intelligence” in context to “thinking-ability”.
The popularity of AI is tremendously growing in this tech- savvy world. Some very interesting AI algorithms are nature inspired such as “Ant Colony Optimization”, “Particle swarm optimization” and Grey wolf optimization” etc.
Ant colony optimization (ACO) is inspired by the release of pheromones by ants and how ants find out the shortest path to return back to their colony while searching food using pheromones.
ACO is used by a variety of AI systems to optimize the cost function and obtain the best performance parameters for a system solving a problem.
Particle swarm optimization (PSO) is based on flock of birds or swarms and how they are always in symmetry while flying and is utilized by a variety of AI applications for optimum co-ordination.
Grey wolf optimization (GWO) is based on a group of wolfs and how they hunt a prey in co-ordination. The GWO is used by AI systems for a strategy based progressive optimization of the cost function.
There are a variety of such interesting nature inspired AI algorithms that make use of natural phenomenon to make machines intelligent.
Machine Learning (ML) is a technology correlated to Artificial Intelligence that is based on the functionality of human brain. The human brain contains an average of 86 billion neurons transmitting electric signals.
The most common ML technology used today, that tries to mimic the functionality of human brain is Artificial neural network.
The term “Machine Learning” consists of two terms “Machine” and “Learning”. Machine Learning may either be supervised, un-supervised or reinforced.
Supervised way of learning:
Supervised way of learning is when a labelled data is used initially for training the machine.
A real-life example of supervised learning may be a father teaching his kid, how to ride a bike. Initially, while training, the father guides his kid until the kid learns to ride.
Similarly, in supervised learning way of ML, at the time of training a system, the machine is provided labelled data and based on the training received by the machine, it performs
Unsupervised way of Learning:
In Unsupervised way of learning, the machine is not given a labelled dataset to be trained, rather the machine tries to identify the relation between the data given at the time of testing on its own and based on the identified relation, classifies the data.
We may take a real-life example of un-supervised learning as an employee of a supermarket when given a task to separate fruits by type, the employee separates the fruits by their kind just by looking on their features without even knowing name of each fruit.
Similarly, a system using unsupervised learning is capable to separate the data into clusters without labelling just by correlating their features.
In case of reinforcement learning, the training method based on rewarding desired behaviours and/or punishing undesired ones, i.e. learning from mistakes.
A real-life example of reinforcement learning may be a kid, learning to ride a bike on own. Initially, while learning he will fall and make mistakes but eventually, will learn and ride the bike.