AI for Retail

CHAPTER 9AI Learning Techniques

For this section, we will pour your AI foundation. Let's start by breaking down AI into the three major AI learning techniques (see Figure 9.1).

Supervised Learning

Supervised learning is the core technique behind most of the examples I gave in the intro where AI revolutionized industries. When you are trying to get self‐driving cars to detect lines on the road or the stop sign or a baby crossing the road, you will be using supervised learning. In speech to text, you will be giving the AI a bunch of sound files paired with captions, which is supervised learning. In translation, you will be giving the AI a bunch of sentences in English and the French translation, for example, as a training pair to teach the AI how to translate between the two languages, so it is supervised learning. In my opinion it is the most useful of the three AI techniques, so I will discuss it the most.

FIGURE 9.1 Breakdown of the 3 major AI learning techniques.

The goal of supervised learning is to learn a function that maps feature vectors (images, text strings, sound files, whatever) to labels (images, probability values, text strings, numbers, whatever), based on example input‐output pairs that are labeled beforehand typically by a human. If the AI model is trained well, the AI will correctly determine the correct output class for any unseen inputs.

For example, ...