Introduction to machine learning

Understand the basics of machine learning

This is a summary for the Intro to machine learning course offered by Google Digital Garage

Artificial intelligence is sometimes called machine learning or machine intelligence.

In the early days, everything was built on logic(Doing mathematical integration problems or playing chess) but the real world is actually very messy, hard logical rules are not the way to solve real-world problems.




In fact, it's probably not one technique for everything, it's probably a bunch of different techniques and combinations of those techniques.

Some applications:

• Anti-lock braking
• Autopilot systems for planes
• Search
• Recommendations
• Maps
• To decide whether or not this particular email is spam or not spam.
• The ability to translate one language to another with your phone

In the last five years, there has been a drastic improvement in what machines are able to do, compared to the previous decade.

With the advent of a lot more data and a lot more computing power, we can change what sort of models we can envision.

Machine learning is all about learning from examples.

The machine could learn from observations about the world, so it can write 500,000 lines of code.

In order to identify patterns, the machine-learning algorithm looks through many examples, so we could generalize from there.

In the task of image recognition, models are trained to take the pixels of an image from which high-level features are learned.

For speech recognition, real-world sounds are mixed into the human voices to recognize them in any noisy environment. So whatever the noise in the environment, the speech recognition systems can understand what you're saying. They can separate out one speaker from another.

That's a promising direction for developing systems that can really navigate the mess of the real world.

The fundamental idea of machine learning is to take some part of a software system that we used to program explicitly with a set of rules and instead, have the machine learn to do that task.

Machine learning is good, both at automating processes and also at making processes more efficient.

Fundamentally it is about the process of automating something that's repetitive and boring to humans. However, now it is operating at a different level. Machine learning can do many things, but it is not necessarily the right answer to every problem. For example: adding two numbers doesn’t need machine learning because we already have software that could that task perfectly.

One of the most important things is having examples of the behavior that you want the machine to learn. So in practice, machines learn best from correct examples being demonstrated to them and then learning to follow suit. So you're looking for an opportunity where you have some task that has been done hundreds or thousands of times, and you have very good records of exactly how it should be done correctly and then you can use machines in order to automate doing that same behavior for millions or billions of times.

Given that machines learn from examples, that they learn from data, you have to have some amount of data to even get started. Machines learn much more slowly than humans do. So whereas it might only take having to show a human how to do something ten times, it would probably take a machine a hundred or a thousand times to learn to do that same task.