Machine Learning is a subdivision and a branch of
Artificial intelligence.
The term Machine Learning appeared with Arthur Samuel, a computer scientist at IBM, around the game of checkers
in the 1950s.
One of the first and successful applications of neural networks (a subset of Machine Learning) happened in Bell
Labs in New Jersey in 1989.
Yann LeCun, Director of Research at Meta and Silver Professor at NYU made an algorithm that recognizes
handwritten digits using the back-propagation network (algorithm for training neural networks).
This is one of the important breakthroughs in AI history. And Yann LeCun is considered as one of the Godfathers
of AI and Deep Learning (A subfield of Machine Learning) today.
The oldest neural network still in use today is the Perceptron (1958). The perceptron is a single layer
neural network, it is also called Linear Binary Classifier because it is used to classify data into two parts.
Machine Learning and statistics are very related
fields. Some even say that they are the same thing.
The major difference between the two is that Statistics emphasizes inference, inference about the relationships
between variables, whereas Machine learning
emphasizes prediction and making the most accurate prediction possible.
Some real-life machine learning examples include :
Product Recommendations, Personalized Ads, Email Spam Filtering, Self-Driving
Cars ...
A Simple Linear Regression is also a machine learning model.
Regression is defined as follows : $$Y = a X +
b$$
\( Y \) is the predicted value and \( X \) is the input data.
In Machine Learning, we call \( a \) as the
Weight and \( b \) as the Bias.
We can use this model if we have a linear problem. For example, we use it in the case of predicting the Height
of a person by their Age.
$$\textrm{Height} = a \ . \textrm{Age} + b$$
If we have a dataset filled with people's heights from when they are kids to when they are adults then we can
predict the height of a given person from their age.
This is a very simple model for doing simple predictions, but most of the time, it's better to start simple
before getting to
more complicated Machine Learning models.
There 4 types of Machine Learning algorithms.
- Supervised Learning
- Unsupervised Learning
- Reinforcement Learning
A Supervised machine learning algorithm requires labeled datasets -labeled inputs and outputs- while an
unsupervised machine learning algorithm doesn't. It can process unlabeled or even raw data.
Classification and Regression are amongst these supervised machine learning algorithms.
For Unsupervised Learning algorithms, we have clustering and dimensionality reduction algorithms (link here!!
-my post).
There is also Semi-Supervised Learning which is the middle ground between Supervised Learning and Unsupervised
Learning. It combines
a small sample of labeled data with another big sample of unlabeled data.
Lastly, Reinforcement Learning.
Reinforcement Learning is based on the idea of reward, and it is the closest thing to how humans learn new
information. The algorithm learns in an interactive environment and tries to maximize reward. it gets feedback
and learns from its past actions and mistakes.
The best way to get into Machine Learning is to work
on projects.
You can start by looking at Kaggle’s courses, they have one for Machine Learning, Python, Deep
Learning and many more…As well as their beginner-friendly competitions like
the Titanic.
Then you can move into making your own projects, researching machine learning topics online and watching
educational Youtube tutorials and videos about them.
For research papers on Machine Learning, Statistics or whatever field you are interested in, you should
check ArXiv.
Machine Learning is an active topic of research and
a very interesting one.
All of your favorite apps use Machine
Learning algorithms and more and more organizations are adopting
machine learning algorithms into their workflows.
If you have any questions or wants to know more about the article,
feel free to email me or
tweet at me!