What's going on everybody! Welcome to my blog! This is my first article and what's better to start with? Let's start by an introduction to Machine Learning. So if you are a beginner in Machine Learning and want to explore this area, you have come to the right article. I will also be publishing various articles on Machine Learning in my blog. So if you like the content, please let me know your views and suggestions on improvements. Please do share this article and also do send your views to let me know how you felt. You may also connect with me over mail or social media and we shall discuss the topics further.

Interested in Predictive Analytics? Then research Artificial Intelligence, Machine Learning, and Deep Learning.

-SupplyChainToday.com

The above line states a lot of terms. It talks about something called "Predictive Analysis". It also tells us to research "Artificial Intelligence" and "Machine Learning". Let us put the Analysis term aside for a while (don't worry, I will come to that shortly), but wait! it says, research Artificial Intelligence and Machine Learning!. But, are they even different?

The answer is, absolutely yes. Both the terms, Artificial Intelligence and Machine Learning are different. How? Let us talk about that.

In computer science, artificial intelligence (AI), sometimes called machine intelligence, is intelligence demonstrated by machines, in contrast to the natural intelligence displayed by humans. Leading AI textbooks define the field as the study of "intelligent agents": any device that perceives its environment and takes actions that maximize its chance of successfully achieving its goals. Colloquially, the term "artificial intelligence" is often used to describe machines (or computers) that mimic "cognitive" functions that humans associate with the human mind, such as "learning" and "problem solving". The word Artificial Intelligence was first coined in 1956, by John McCarthy (no, I am not good with history. This was an excerpt from Wikipedia. Credits to Wikipedia!). So, basically, the goal in Artificial Intelligence is to simulate natural intelligence to solve complex problems and the aim is to increase chance of success and not accuracy. Thus, Artificial Intelligence is Decision Making.

While, Machine Learning is the learning in which machine can learn by its own without being explicitly programmed. It is an application of AI that provide system the ability to automatically learn and improve from experience. In Machine Learning, the device does not "make decisions to solve problems" but it merely "learns from provided data of a specific task" and "optimises its performance on that particular task" without being explicitly programmed. Thus, Machine Learning allows system to learn new things from data. Here, the main aim is to increase accuracy. Thus, in a nutshell, Machine Learning allows system to learn new things from data.

So, this article mainly aims to talk about Machine Learning and give an introduction about it. So, as discussed, Machine Learning, abbreviated as ML, aims to perform accurately in a given task by learning from provided data. And, when we are learning Machine Learning, what do we learn exactly? Essentially, we learn about various ways, technically, algorithms to "teach" the devices how to perform in a given task. So, what type of tasks can the device "learn" about? Here is a list of (most important) tasks for which we use Machine Learning to "teach" the devices:

1. Predicting something (remember we came across something called "Predictive Analysis"?)
2. Classifying input into Binary or Multiple groups
3. Finding out similarities between various data objects

Now, these being the most implemented tasks in Machine Learning, how to implement them? There are basically three main types of approaches of teaching a computer or a device to learn itself:

1. Supervised Learning
2. Unsupervised Learning
3. Reinforcement Learning

So, let us talk about each of the above methods:

1) Supervised Learning: This is a relatively common approach when we compare the usage of all the three approaches. Let me explain this approach with the help of an example. Suppose, we have just joined school and it is our very first mathematics class. Our teacher, starts to teach us. And our first topic is Addition of two single digit numbers. So, think of it yourself. How will the teacher teach us? Exactly. He/She will first write down some 'example questions', show us how to solve and give us the answers. Then, he/she will give us some "practice problems" without the answers which we will solve according to what the teacher taught us and we will then check our answers with the actual answers. If we are correct, that means we have "learnt" addition. If not, the teacher will give us some more examples untill we learn it completely. In the same way, in this method, we will be providing our device with example inputs along with their outputs. The device will try to figure out how to get the output from the input, based on the algorithm we use, and will learn to perform the task. Next time, if we will give it a completely unseen input, it will try to find its output based on the learning it has done previously. This is why, it is called "Supervised", we are supervising the learning, as a teacher does.

2) Unsupervised Learning: This is another interesting approach on solving a Machine Learning problem. This is a little bit different from Supervised Learning. Just as the name suggests, Unsupervised Learning aims to teach the device to perform the task without a dedicated supervision. In the previous method, we provided the device some input and the corresponding outputs to learn from. Here, we will just be providing the device with the inputs. The device, according to the algorithm used, will try to find similarity between the data. Thus, this approach is usually used for clustering problems. Given a set of inputs, the device will classify all the inputs into different clusters according to the similarity in their properties and when given a new input, it will assign the new input its respective cluster. And as you may have already guessed, the algorithms used in this approach are different than the algorithms used in Supervised Learning

3) Reinforcement Learning: This is another approach of Machine Learning. Let me tell you about this approach with the help of an example. Suppose, you have recently adopted a pet. Now, suppose you are training your pet to sit down when you say "sit". So what would be your approach? You would speak to the dog - "Sit!". And if it sits? You will give it a treat. And if not? It gets nothing. So what will the dog do? Whenever you say "Sit!", it will promptly sit down. Why? Because it knows that by doing something like this, he is able to earn a reward. Similar to this, we try to train the device based on a reward and penalty method. We instruct the device to try and reach to a goal state from the given state. Each step it takes has either a reward or a penalty associated with it. If the device takes a wrong step, it will be penalised and based on this penalty, it will learn that it was a mistake. If it gets a reward, it will understand that it was a correct step and try to take similar steps for some further similar type of problems.

So, this was a very short introduction about Machine Learning and its approaches. Please let me know your feedback regarding this article and if you liked the article, please share it using the links provided below. Also, do suggest me about any improvements which I may implement to make the blog even better. Further, I will be writing about various Machine Learning algorithms in future posts. Please stay tuned! Until then, have a great time!