Machine Learning basics
Posted on December 5, 2024 • 3 minutes • 621 words
We have heard about Artificial Intelligence, Machine Learning and Deep Learning topics a lot. Everyone is talking about these topics. Machine learning seemed very interesting topic to get into, I know I am late already but “Better late than never 😁”.
I am starting on these topic, so I thought it would be amazing to share with you, my learnings and hope that will help you too. So, Let start with what comes inside what.
Here is the hierarchy of Artificial Intelligence, Machine Learning and Deep Learning in visual:
I am going to share the basics of Machine learning in this blog, lets understand ML better.
Lets go to the history first
Do you AI was born in 1950s when a few pioneers from nascent field of computer science started asking if they can make a computer “think”, a question which we are still exploring today.
That time, most experts believed that human-level AI could be achieved by having programmers handcraft a sufficiently large set of explicit rules for manipulating knowledge stored in explicit databases, the approach was called symbolic AI. It dominated its way from 1950s to 1980s and reached its popularity peak during expert systems boom of the 1980s.
Although symbolic AI was amazing to solve well-defined, logical problems such as playing chess, it turned out to fairly complex for solving more complicated, fuzzy problems such as image classification, speech recognition and natural language translation. Thats when a new approach called Machine Learning arose and took place of symbolic AI.
What is Machine Learning?
We all know how the classic programming works, we write a set of instructions, we give inputs and we expects the results. But machine learning is different: the machine looks at the input data and the corresponding results, and figures out what the rules should be. A machine learning is trained rather than explicitly programmed.
Its presented with multiple examples relevant to a task, and it finds statistical structure in these examples that allows the systems to come up with rules for automating the task. Example, if you give multiple pictures, tagged with people’s name, then system would learn statistical rules for associating specific pictures to specific tags.
Actually there are total 3 inputs for Machine Learning:
- Input data points : such as large set of images
- Examples of expected output: such as tags on images like ‘cat’, ‘dog’ etc.
- A way to measure if the algorithm is doing well : its important to identify the different / distance between current output and expected output. These feedbacks are needed to adjust the way the algorithm works. This adjustment step is called learning.
Representation is an important part of Machine learning, its a different way to look at data. For example, color image can be represented in RGB(Red-Green-Blue) and HSV(Hue-Saturation-Value), they are different representation of same data. Some tasks may be difficult with one form and easier with another.
Machine learning models are all about finding appropriate representations for their input data - transformations of the data that makes it more easier to handle task in hand.
Example:
With this representation, the dark brown and light brown classification problem can be expressed as simple rule: “Dark points are such that x>0” or “Light points are such that x<0”. This new representation combined with this simple rule, neatly solves the classification problem.
So we can say, Machine learning is searching for useful representations and rules over some input data, within a predefined space of possibilities, using guidance from a feedback signal.
Machine learning is fun and I hope you understood the basis of it. I will bring more examples and we will next get into Deep learning, so stay tuned and see you in next blog. 👋