How Machines Learning Models Learn

A nice example from the Machine Learning for Web Developers (Web ML) video course (47 videos) by Jason Mayes:


A machine learning model is essentially an algorithm that transforms numerical inputs into numerical outputs used in applications. 

On the left is a magnified image of a person.

Now, as a human, you can probably make out where the face is pretty fast. But a computer cannot see objects like you and I can.

It just sees numbers, as shown by the middle image. Here, numbers from 0 to 255 represent different shades of gray.

And on the right-hand side is what would be fed into your Machine Learning model as an input.

Do you still see the face in that?

Well, as a human, it's nearly impossible to recognize in this form. But this is what your Machine Learning model must learn from and some how take these numbers to produce some output that represent how sure it is that a face is in those input numbers.

And this is why the quality of training data is very important. Generally speaking, the more diverse and well labeled your data, the better Machine Learning model you will be able to train that can generalize well to unseen examples in the future.

All machine learning models are initially untrained.

The same YouTube course is available with additional resources & quizzes through EdX

Comments

Popular posts from this blog

Maven Crash Course - Learn Power Query, Power Pivot & DAX in 15 Minutes

"Data Prep & Exploratory Data Analysis" course by Maven Analytics

Oracle Cloud Infrastructure 2024 Generative AI Professional Course & Certification Exam (1Z0-1127-24)