Machine Learning – Basic Introduction for Beginners
Have you ever imagined how can an e-commerce site know which products you are more likely to buy next? How it can suggest you which are the other videos in YouTube you are interested in watching? How camera click an image automatically when it detects a smile on your face (A Computer Vision Problem)? By what method can a machine is able to detect Real Vs Fake Smile? All these amazing things use the concept of MACHINE LEARNING. Commonly abbreviated as ML.
It is the branch of computer science which gives the computer the ability to learn without being explicitly programmed (Arthur Samuel). Prior programmers used to write tons of lines of code just to perform a simple task. With this technology, we are now able to teach the machines how to learn from experiences.
So what exactly do we understand by the term Machine Learning?
Machine Learning algorithm gives a machine the ability to find out the hidden patterns that are present in the dataset. It learns its behavior using Statistics behind it.Later with these discovered patterns do the task accordingly.
Why Machine Learning is Gaining the LIMELIGHT?
In the near future, every object in this universe will be throwing out data. It includes our phones, internet searches, cars even our body. Everything on this Earth is generating tons of data in every single minute or even less. Every single thing we do, we are leaving our digital traits in the Universe. It can be our location or our buying history or anything. It seems as if everything is recording in real-time.
So now it is very important to use this massive amount of dataset for the well-being of society and for the individual. For this, we need some techniques which can quickly process this massive amount of dataset and make some insights out of it.
Here comes the use of Machine Learning because such complex problems are impossible to solve by hand. Therefore, we need some way where a machine can automatically learn from these data and find some patterns and from that give some kind of meaning to that data.
Types of Machine learning Algorithm
Machine Learning Algorithm is broadly classified into 4 categories:
- Reinforcement learning Figure 1: Classification of Machine Learning Algorithm
(Source of image: http://www.isaziconsulting.co.za/machinelearning.html)
Supervised Learning Algorithm:
In this type of algorithm, we first train the model by providing a set of inputs with its corresponding labeled correct outputs. After that, we make a generalization function. This function helps in mapping input with its corresponding correct output.
How it works
Example: Have you ever wondered how our e-mails automatically filter out the important mail from the Spam email? The algorithm which works behind this job comes under supervised learning.
We train the model with the set of words which are frequently used in the mail. Training involves specifying which words indicates “it’s a “SPAM” or Not. From this, the model will make a generalized function. This function helps in detecting whether it’s a Spam or not. When real-time datasets are provided to this model, it can easily detect by analyzing the words present in the mail whether it is a Spam or not.
Still, it needs further investigation like whether there is any overfitting or underfitting. These terms are out of the scope of this article. The supervised learning algorithm is preferred where we already have some datasets with its corresponding correct label. Then based on training we can make a generalized model which will make a prediction for the new unlabeled dataset.
Figure:2 How Supervised Learning Algorithm work
Unsupervised Learning Algorithm:
In this type of algorithm, we only have a set of inputs. Here our goal is NOT to make any prediction instead we are trying to find out patterns which exist in the data. The target of this algorithm is to find the similarity among the dataset. Here, we try to find out the structure of the data so that we can create sub-groups out of it which are similar in some sense. The main goal of this algorithm is to know more about the data by finding the similarity among the dataset.
Example: This type of algorithm is used where we want to group the data based on some specific patterns.Have you ever noticed how Facebook suggests you “Friends You may know?”.This thing is performed using “Clustering”.This is one of the famous technique in Unsupervised learning.It uses the same idea of clustering and finds friends which are similar to you in some way like the location where you are currently living or the college in which you have studied or they may be your friends of friends.Through this, it will create a sub-groups of persons which matches with you and suggest you that friends which falls in that cluster.
Figure:3 Unsupervised Learning Example
Semi-Supervised Learning Algorithm:
As the name suggests Semi-Supervised learning lies between Supervised and Unsupervised learning because it trains the model with the large number unlabelled data and a very small labeled dataset. Many real –time problem falls into this category because its very time consuming to label the dataset and the unlabeled dataset are very cheap and can be easily available. In the real time, not every data which we get is labeled.
So there we use this type of algorithm to solve our problem.
This algorithm is a reward based type. Here the machine interacts with the real-world. It performs various tasks for which if the task is performed correctly the machine is rewarded or given positive reviews and if it does some mistakes it is given negative reviews by this way it trains itself.
Example: – The famous example for this type of algorithm is the Self-Driving car which learns how to drive by interacting with the real-time environment.
Some Interesting Application of Machine Learning:-
In recent times everyone is using Machine learning to reshape their business for boosting their efficiency.
Amazon Go is a brick-and-mortar grocery store.”Brick-mortar” means something which has a physical presence. Amazon is building stores where a customer will just grab the item which they want to buy and leave the store. It automatically deducts the money from their Amazon account whenever one picks any item to buy. No carrying of cash or standing in the queue is required.
Face detection: The face detection feature in mobile cameras is an example of what machine learning can do. Cameras can automatically snap a photo more accurately when someone smiles. This is because of the advances in the machine learning algorithms. Even it can collect and cluster similar kind of pictures to make some interesting story out of it.
Medical diagnostics can now able to detect a particular disease prior to the period when the actual disease will occur. For the treatment of cancer, we are now taking the help from IBM Watson. It gives personalize treatment for individual cancer patients by analysing the medical records of patients. Many companies have joined hand to fight against cancer by using Artificial Intelligence. Recently former Vice President of United States of America, Joe Biden made a promise to cure cancer using Artificial Intelligence. This initiative is joined by many companies.
This technology can easily monitor our Health. Many Applications are there which can easily tell you when you are going to be feeling depressed next time.
Recommendation Systems: E-commerce sites uses machine learning techniques to generate a recommendation for the users to buy a specific product. It can easily detect what are the other products you are more likely to buy.
This article covers a brief introduction to Machine learning, its basic classification of algorithms and some of its applications. Still, we need to cover a long distance to reveal the application of ML. Daily Researches are going on in this field and the objective is to make efficient algorithms.Thus, We need powerful Algorithms which can handle such large datasets. The Future has lot more patterns to detect in this area.