Artificial Intelligence in Practice

Artificial Intelligence in Practice

DOI: 10.4018/978-1-7998-2036-9.ch005

Abstract

In this chapter, the authors discuss machine learning techniques and artificial intelligence applications, their role in business, and present a practical application of it. They try to highlight how important machine learning can be in data-driven organisations, where the cost and/or the advantages to implement such tools are far greater than having a human—or a team of humans—doing it.
Chapter Preview
Top

Machines And Their Learning

In the previous chapters, we have discussed about the importance of data and its proper use. We have also introduced artificial intelligence, and the marvellous achievements we could already celebrate. In this perspective, machine learning can be thought of as a set of mathematical rules - also known as algorithms - to analyse data. But how does machine learning work? There are several techniques that can be used to implement machine learning, according to the type of learning programmers have decided to use.

Thus, we can start with the most common technique in the field, also known as supervised learning, whose main objective is to train the machine into recognising one or more elements that have been previously identified, as contained in a digital data stream (which can be an image, a face, a sound, and so on). Ideally, supervised learning takes place when either we are trying to predict the class of an item, using what is known as classification models, or a number, in which case we are exploiting regression models. In both cases, we could use simple linear1 and logistic regression2 or more complex decision tree families3. Furthermore, this implies that we need to build a training set with thousands or millions of images, which will be supervised by analysts to make sure the right data are inserted, as well as to make sure the machine is actually learning and correct its errors, in case any have been detected. Once the first stage of the training process - also known as learning phase - has been completed, new images are introduced into the program, to see if the learning actually took place and it is sustained even with other tons of images added to the mix4. This process has benefitted from advancements in computational power and availability of data from the web and the Internet of Things, in particular when it comes to translation of text (Google Translate is a powerful AI tool that is able to find patterns and grammatical structures to translate words and more complex text), as well as facial recognition systems (see for example, Zakharov, Shysheya, Burkov & Lempitsky, 2019). Interestingly, supervised learning can be enhanced through a reward mechanism, according to which the model provides its margin of error, once it has understood the correlation between inputs and outputs and has been able to identify the probability of success in reaching specific objectives.

Complete Chapter List

Search this Book:
Reset