Abstract
In the previous chapter we developed an IoT application that will ingest temperature and humidity data. This data will be processed by an ML service and a prediction will be made: is it likely to rain or not?
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
In the new covid19 world we find ourselves in, perhaps someone will build a machine learning system that is fed data from a bank of sensors that read a variety of personal health-related conditions (perhaps IR temperature, particulates, and C02 in exhaled breath, irregularity of heartbeat, frequency and pitch of coughing; there are so many different sensors available). The ML service would be trained to predict whether someone is infected with the virus, allowing us to identify the people most in need of early help.
- 2.
These are the same variable names that we saw in Chapter 11 when we were setting up the IoT application. The names originated from the ML service we used, and throughout the process of setting up the IoT application, we did not need to enter them at all. There are a few places in the Azure workflow where you can view the variable names – for example, check Chapter 11, Section 11.7, Step 1. It is essential that you spell the variables correctly and use exactly the same case as is used in the IoT application.
- 3.
If a consumer ML service allows members of the public to train it, then there is a risk that people will mistrain it, either by accident or willfully. Sometimes getting the public to train an ML is a good idea:
-
Google’s Captcha software purportedly identified humans from bots trying to access a website. The responses we have all provided when using Captcha are used to train Google’s ML image recognition software.
It has also been known to backfire:
-
In 2016 Microsoft released a chatbot called Tay on twitter. Tay was set up to learn from conversations. Within a day Tay had developed a nasty streak that resulted in Microsoft having to suspend its account. This was an example of willful mistraining of an AI and shows that caution should be exercised when training an ML service.
-
Author information
Authors and Affiliations
Rights and permissions
Copyright information
© 2020 Philip Meitiner, Pradeeka Seneviratne
About this chapter
Cite this chapter
Meitiner, P., Seneviratne, P. (2020). Connecting an Edge Device to the IoT Application. In: Beginning Data Science, IoT, and AI on Single Board Computers. Apress, Berkeley, CA. https://doi.org/10.1007/978-1-4842-5766-1_12
Download citation
DOI: https://doi.org/10.1007/978-1-4842-5766-1_12
Published:
Publisher Name: Apress, Berkeley, CA
Print ISBN: 978-1-4842-5765-4
Online ISBN: 978-1-4842-5766-1
eBook Packages: Professional and Applied ComputingApress Access BooksProfessional and Applied Computing (R0)