Understanding Machine Learning
Machine learning is a subset of artificial intelligence (AI) that involves
training algorithms to learn from data, without being explicitly programmed.
In other words, instead of being given a set of rules to follow, the machine
learning algorithm is trained on a large dataset and learns to identify
patterns and relationships within the data on its own.
The process of machine learning involves several steps. First, the
algorithm is trained on a labeled dataset, which means that the correct
output is provided for each input. For example, a machine learning algorithm
that is trained to recognize images of cats might be given a dataset of
thousands of images, with each image labeled as either a cat or not a cat.
Once the algorithm is trained, it can be used to make predictions or
decisions on new, unlabeled data. For example, the cat recognition algorithm
could be used to identify cats in a new set of images.
Machine
learning (ML) technology is a machine that was developed to be able to learn by
itself without direction from the user. Machine learning builds on other
disciplines such as statistics, mathematics and data mining so that machines
can learn by analyzing data without needing to be reprogrammed or instructed. In
this case machine learning has the ability to obtain existing data with its own
commands. ML can also study existing data and the data it obtains so that it
can perform certain tasks. The tasks that ML can perform are very diverse,
depending on what they learn.
The
term machine learning was first put forward by several mathematical scientists
such as Adrien Marie Legendre, Thomas Bayes and Andrey Markov in the 1920s by
stating the basics of machine learning and its concepts. Since then, ML has
developed a lot. One example of a fairly well-known application of ML is Deep
Blue which was made by IBM in 1996. Deep
Blue is a machine learning developed to be able to learn and play chess. Deep
Blue has also been tested by playing chess against a professional chess
champion and Deep Blue managed to win the chess match. The
role of machine learning helps humans in many fields. Even now, you can easily
find the application of ML in everyday life. For example, when you use the face
unlock feature to open your smartphone device, or when you browse the internet
or social media, you will often be presented with several advertisements. The
advertisements that appear are also the result of ML processing which will
provide advertisements according to your personality.
Actually there are
many examples of the application of machine learning that you often encounter.
Then the question is, how can ML learn? ML can learn and analyze data based on
data provided at the start of development and data when ML is already in use.
ML will work according to the technique or method used during development. What
are the techniques? Let's see together.
There are three main types of machine learning: supervised learning,
unsupervised learning, and reinforcement learning.
Supervised Learning
Supervised learning
involves training the algorithm on a labeled dataset, as described above.
The algorithm learns to identify patterns in the data and can then make
predictions or decisions based on new, unlabeled data. Supervised
learning techniques are techniques that you can apply to machine learning where
you can receive information that is already in the data by giving it a certain
label. It is hoped that this technique can provide a target for the output that
is carried out by comparing past learning experiences.
Suppose you have a
number of films that you have labeled with a certain category. You also have
films in the comedy category including the films 21 Jump Street and Jumanji.
Besides that, you also have other categories, for example the horror film
category like The Conjuring and It. When you buy a new film, you will identify
the genre and content of the film. After the film is identified, then you will
save the film in the appropriate category.
Unsupervised Learning
Unsupervised learning
involves training the algorithm on an unlabeled dataset, where there is no
correct output provided. The algorithm learns to identify patterns and
relationships in the data on its own, without any guidance from a human
expert. Unsupervised
learning techniques are techniques that you can apply to machine learning that
are used on data that does not have information that can be applied directly.
It is hoped that this technique can help find hidden structures or patterns in
data that do not have labels. Slightly
different from supervised learning, you don't have any data that will be used
as a reference beforehand. Suppose you have never bought a movie at all, but at
one time, you bought a number of movies and wanted to divide them into several
categories so they were easy to find.
Of course, you will
identify which films are similar. In this case, suppose you identify based on
the genre of the film. For example, if you have the Conjuring film, then you
will save the Conjuring film in the horror film category.
Reinforcement learning
Reinforcement learning involves training the algorithm to interact
with an environment and receive feedback in the form of rewards or
punishments. The algorithm learns to maximize its rewards by learning which
actions lead to positive outcomes.
How
Machine Learning Works
The
way machine learning actually works varies according to what kind of learning
technique or method you use in ML. But basically the principles of how machine
learning works are still the same, including data collection, data exploration,
model or technique selection, providing training on the selected model and
evaluating the results of ML. To understand how ML works, let's review how some
of its implementations work below.
AlphaGo
is a machine learning developed by Google. When it was first developed AlphaGO
would be trained by providing 100 thousand Go match data for him to study.
After AlphaGo has the provision and knowledge of ways and strategies to play
the Go game from studying 100 thousand Go match data. AlphaGo will learn again
by playing Go with himself and every time he loses he will improve the way he
plays and this playing process will be repeated up to a million times.
Improvements
in how to play AlphaGo were made by himself based on his experience when he
played against himself or against other people. AlphaGo can also simulate
several matches at one time simultaneously. This means that at one time he can
do several Go matches at once to learn. So that the learning process and
experience playing Go can also be more numerous than humans. This was proven
when AlphaGo played with the world Go champion in 2016 and he could be the
winner.
From
the application of machine learning to AlphaGo, we can understand that machine
learning will continue to learn as long as it is used. Just like the face
detection feature on photos owned by Facebook, it will learn to recognize your
facial pattern based on the sign you enter when posting a photo. From the
person you tagged in the photo, ML will make this information a medium for
learning.
So don't be surprised
if machine learning is often used, then the level of accuracy is getting better
than at the beginning. This is because machine learning has learned a lot over
time from users' use of machine learning. As with Facebook's face detection
feature, more and more people are using this feature and tagging people in
photos, the level of accuracy of people detected is getting better.Machine learning has a wide range of applications, from image and
speech recognition to fraud detection and recommendation systems. As the
amount of data generated by businesses and individuals continues to grow,
machine learning is becoming increasingly important for extracting insights
and making decisions from this data.
Comments