The Parts Part 0

As you know, AI has lots of subcategories, and Machine Learning has lots of subcategories. The technical style of machine learning we'll be using, is the one most people think of and know. As it happens, it's also the most business applicable as well.

JARGON ALERT! We'll be looking at discriminative computer vision models in supervised learning. Or as most people think of it, an AI that can identify stuff from photos.

This is what identifies cancer in X-rays, people on doorbell cameras, and even famous actors on your smart TV.

Another aspect of this course is that we'll be focusing on the client side and utilizing TensorFlow.js. TensorFlow is Google's framework for Machine Learning. It handles a lot of lower level math so that developers can focus on higher level Machine Learning applications. TensorFlow (no JS) is meant for use in Python because that's where the majority of Machine Learning enthusiasts exist today. As you might already know, Python is a popular language for Data Science, because Python gives a nice balance between readability and speed. Lots of business and training courses teach a mix JavaScript with Python to create a service that exists on the server, but is accessed by the client. Keeping the model processing on the server means you're paying for it, and sometimes that can hold you back. Take a look at selfie2anime.com, which had to add a "high demand" message.

If the implementation was on the client, their solution would be far more scalable and measurably private with user photos. It's common to see machine learning models kept on the server, and because of that most stop at TensorFlow and forget about TensorFlow.js, but not us!

While there are lots of benefits of keeping models on the server, it's great for a business built around AI, but most products have to go to the user. TensorFlow.js is also by Google, and has a growing API that brings functionality of TensorFlow to a broad spectrum of platforms via JavaScript. That's why we chose to do this course in TensorFlow.js. It's a fantastic way to leverage machine learning everywhere.

Moving the model 100% to the client, allows you to run machine learning offline, privately, and with static server costs. TensorFlow.js runs completely in the user's browser, removing round-trip delays, while you the developer enjoy the simplicity and universality of keeping everything in JavaScript. There's even availability for TensorFlow.js in React Native, which gives you the ability to expand your TensorFlow.js clients to all the platforms of mobile and beyond!

With all that said, if you ever require your model behind a server, there's also Node TensorFlow.js support, which gives you the ability to apply everything you learn in this course to server-side JavaScript API. This course is going to teach you with a framework of Machine Learning that you can take everywhere.