Hi, I’m Machine Learning! Definitely, I Can Learn
What is the first thing you think of when you hear the word “Machine Learning”? Yes, not only humans, machines can also learn. So what do machines learn, how do machines learn, and what are the results? To answer these questions, let’s get acquainted with Machine Learning.
Jason Bell, in his book Machine Learning: Hands-On for Developers and Technical Professionals says:
“Machine learning is a branch of artificial intelligence. Using computing, we design systems that can learn from data in manner of being trained. The systems might learn and improve with experience, and with time, refine a model that can be used to predict outcomes of questions based on the previous learning.”
Machine Learning is a branch of Artificial intelligence that studies how a machine/system can learn from the data given to “them”. A system can be said to have implemented the Machine Learning method when it can learn from data and can provide a prediction based on its learning outcomes.
What does Machine Learning learn?
If humans can learn from books, then Machine Learning can learn from data. One of the factors that affect the “intelligence” of Machine Learning is the data provided to “them”.
Lots of data given to learn is good, but more data does not guarantee the “intelligence” of a Machine Learning. Good data quality will be more helpful. Just like humans, reading a lot of books doesn’t guarantee to make us smarter unless the books are of good quality, right?
How does Machine Learning learn?
Kevin P. Murphy, in his book entitled Machine learning: A Probabilistic Perspective says:
“In particular, we define machine learning as a set of methods that can automatically detect patterns in data, and then use the uncovered patterns to predict future data, or to perform other kinds of decision making under uncertainty (such as planning how to collect more data).”
Basically, Machine Learning learns by detecting patterns in data and then using those patterns to provide predictions.
Learning methods in Machine Learning include:
a) Supervised Learning
In the Supervised Learning method, the terms training data and test data are known. Training data is data that already has a known label or result. This method is called Supervised Learning because Machine Learning will be guided by the labels on the training data. While the test data is new data that does not have a label.
In this method, Machine Learning will first learn the training data and detect patterns in the data using an algorithm. This stage is also called Model Development or Model Training. Where algorithms that are commonly applied to Supervised Learning include: Naive Bayes, SVM (Support Vector Machine), Neural Networks, Random Forests and others.
After learning, it’s exam time! At this stage, the Machine Learning Model is ready with the patterns it has found. Then, it will be given input test data that must be predicted.
The following is an example of Supervised Learning in the classification process. For example, there is data in the form of various pictures of flowers that have been labeled “Rose”, “Suns”, and “Lilies” as training data. Machine Learning will learn the training data and find patterns in the data, so it can classify the given test data as a “Roses”.
b) Unsupervised Learning
Just like Supervised Learning, in this method Machine Learning will also learn data and detect patterns using algorithms.
The difference is that in Unsupervised Learning, the data learned is not known by its label. Machine Learning is required to find similarities in each data and provide predictions. Common algorithms applied to Unsupervised Learning include: K-Means, Hierarchical Clustering, and others.
The following is an example of implementing Unsupervised Learning in the clustering process. For example, clustering will be carried out on various flower images. Machine Learning may classify the data as types “X”, “Y”, and “Z” or as “Stemmed” and “Unstemmed” depending on what pattern Machine Learning finds from the flower images.
In everyday life, we can find lots of Machine Learning applications around us. Some of them are:
1. Image Recognition
When your smartphone can distinguish your face from the faces of your friends and then group them in different folders in the gallery, that’s Machine Learning!
2. Chatbots
Your chat is not answered by him/her? Don’t be sad, try chatting with a chatbot. The Machine Learning method is applied to the chatbot, so it can answer your questions.
3. Voice Recognition
Another example of implementing Machine Learning is Voice Recognition. If you are tired of chatting with chatbots, you can try chatting directly with virtual assistants such as Google Assistants, Siri, Alexa, Cortana, and others.
4. Recommendations for you
Various websites or applications use the Machine Learning method to learn user patterns, so that they can provide product recommendations, content or promotions according to the characteristics of each user.
For example, on an E-Commerce site, you often see briefcase products. Then product recommendations or promotions for you will be about briefcases.
5. Detect Spam in E-mail
Have you ever thought how a spam e-mail can be detected? Yes, that is the job of Machine Learning.
So how? Now are you familiar with Machine Learning? Interested to find out more?
Original article published on Medium by Rianita Giovanni Katryn, June 12, 2020.
https://medium.com/mandiri-engineering/hai-aku-machine-learning-mesin-yang-bisa-belajar-f6e8e858dd04
