MACHINE LEARNING IS FUTURE

What is Machine Learning

Machine Learning is a subfield of artificial intelligence.

Artificial Intelligence is the umbrella in which machine learning comes.

Machine learning in simple words means feeding data to the machine so that it can learn and do tasks accordingly. Through machine learning, we can learn various things based on the inference of the large amount of quality data provided to it about the world that we cannot possibly as human beings can study or appreciate. So, machine learning is when we train computers to understand patterns by taking a look at instances in data and recognizing those patterns, and therefore applying them in new things they haven’t seen before.

How does Machine Learning work?

In traditional programming exclusive programs are written and inputted in the computer and then data is taken and the appropriate output is produced.

An example of it would be a square root finder.

But in the Machine Learning approach, the output is given to the computer first. The examples are given to the machine that we want the program to do, that is labels on data, characterization of different classes of things. From that characterization of output and data, the machine learning algorithm will then create a program through which we can infer new information about things.

And that can create a nice loop, that is the machine learning algorithms will learn the program and we can use it to solve various other problems.

So how can we learn or how can a computer learn?

So, for us as human beings, there are a couple of possibilities. Memorizing facts, the boring one. This accumulation of individual facts is known as declarative knowledge.

They are limited by time to observe facts and memory to store facts.

A better way to learn is to deduce new information from the old, that is, generalization or also known as imperative knowledge.

Limited by the accuracy of the deduction process, predictive activity assumes that the past predicts the future.

In the first case, we built that in when we wrote that program to do square roots. But what we would like in a learning algorithm is to have much more of that generalization idea. We are interested in extending our capabilities to write programs that can infer useful information from implicit patterns in the data. So not something explicitly built, like the comparisons of weights and displacements but implicit patterns in data and have that algorithm figure out what those patterns are, and use those data to create a program from which we can infer new data about objects about spring displacements, etc.

In broad ways machine learning can be divided into the following:

Supervised learning

Unsupervised learning

Reinforcement learning

Supervised learning: In this case, every new example that is given as training data has a label on it. Now we try to find a rule that predicts the label associated with a previously unseen input based on those examples.

It is called supervised because we have a label associated with every example.

Example: Cricket players, labeled by position, weight, and height.

Unsupervised learning: In this case, there is no label associated with the various examples taken. The goal is to find natural ways to group these examples.

Reinforcement learning: In this case, the learning is the replica of how humans learn from their environment. The algorithm learns from themselves using the trial-and-error method. Favorable outputs are encouraged and unfavorable outputs are discouraged. 

Why Machine Learning?

We teach machines to learn from data to build a model from the data or a representation of that to make a prediction. One of the places we often find machine learning in the real world is in the things like recommendation systems

For example, Google, Facebook, YouTube, Spotify, and others use machine learning for recommendation and for targeting ads.

A machine learning algorithm can analyze millions of data which is not possible for a human to do at a given rate. Machines are great at predicting based on what they have seen in the past but they are not creative.

Another example of machine learning would email spam filtering model. 

How can we improve machine learning algorithms

It is fairly new that we can solve all these problems and start to build these products and apply them in the businesses. And so it is an ongoing developing process. Our world is gradually evolving to become more technologically reliant. And one of those technologies that are revolutionizing society is machine learning. Every part of the world is using machine learning be it the simplest or the most sophisticated ones.