Introduction

During this lecture, we will provide a general overview of how a machine perceives data and learns from it. As an initial example, we will train a simple regression model -machine- to estimate the conversion from kilometers to miles. The training process will first be introduced using an iterative method based on basic error metrics. Next, we will present a deterministic approach for the same task. Finally, we will explore a basic linear classification machine and its general training process.

Your first predicting machine

Let us start by proposing a basic Machine that can process information for us. We provide some input data, and the machine “thinks” in order to generate the corresponding output. From now on, we will refer to each piece of input data as an instance.

The machine can “learn” about a process by analyzing the relationship between its inputs and outputs. In this AI context, learning means adjusting the model’s parameters so that the predicted outputs become closer to the real/measured ones.

Thus, our first model proposal is a simple one with just one parameter:

$\hat{y} = \theta x$

Here,

• $\hat{y}$ is the model
• $\theta$ is the unique parameter model
• $x$ is the input data