Little Bits of Artificial Intelligence

1. What is Learning with Labeled and Non-Labeled Data?

Imagine we’re in a classroom, and our teacher gives us two types of papers:

• Labeled papers: These papers have both pictures and names. Like a picture of a cat with the word “Cat” written under it.
• Unlabeled papers: These only have the pictures — no names written on them.

2. What is Semi-Supervised Learning?

Semi-supervised learning

is like a smart student who:

1. Looks at the labeled pictures and learns what a Cat, Dog, and Apple look like.
2. Then, looks at the unlabeled pictures and says,
“Hmm… I think this one looks like the labeled cat I saw earlier. So, this must be a Cat too!”

This student doesn’t need the teacher to label everything. They guess the labels by learning from a few examples.

3. Why Use Semi-Supervised Learning?

Because labeling everything takes time!

Imagine if we had 10,000 pictures, but only 100 were labeled. Instead of labeling all 10,000, the computer can learn from the 100 labeled ones and make smart guesses for the rest.

4. Simple Practical Examples

1. Toy Sorting Game

We have a few toy pictures labeled:
“Teddy Bear”, “Toy Car”, “Yo-Yo”
Then we get lots of pictures with no names.
We look at the labeled ones to guess the names of the rest.

2. School Lunch Tray Example

• The school gives us some trays where they wrote what’s on them: (Pizza), (Apple), (Milk)
• Then they give us many trays without names.
• We look and say, “This tray looks like the one with pizza — it must be too!”

3. Voice Assistant Learning

• A voice assistant like Siri hears a few commands that are labeled:
  “Turn on the lights” , “Play music”.
• Then it hears new ones without labels and guesses what the user means based on what it has already learned.

5. In Short: