CEC Semenster Undefined
Analytics for the IoT: A Deep Dive into Algorithms
Louis Giokas -
September 12,
2016
Introduction to Analytics for the IoT
ToDay we will present an overview of the types of statistical techniques useful for the IoT and how statistics is used in this context. Some differences with standard statistics will be highlighted. The requirements for IoT analytics will also be presented. Our intent is to discuss the algorithms in the context of engineering applications in the IoT.
Course Resources
Special Educational Materials
UDOO Neo Ideal for IIoT Designs Where Extensive I/O Sensors and Powerful Compute Resources are KeyHow a New Breed of Wireless MCUs and Modules Are Enabling Connected IoT ApplicationsDevelop Powerful Event-Driven Video
September 13,
2016
Languages and Environments
To implement the analytics we are discussing this week, one needs software and a platform to run it on. We will discuss languages such as R and Python and more conventional languages and traditional statistical processing systems. Because the size of the data is often massive and growing, platforms become important. It is key to pick one that can start small and grow in power without having to completely rewrite the application. We will give an overview of platforms specific to the IoT.
Course Resources
Special Educational Materials
UDOO Neo Ideal for IIoT Designs Where Extensive I/O Sensors and Powerful Compute Resources are KeyHow a New Breed of Wireless MCUs and Modules Are Enabling Connected IoT ApplicationsDevelop Powerful Event-Driven Video
September 14,
2016
Classification
A major area of machine learning concerns classification. We will describe both supervised and unsupervised learning techniques and discuss when and how to use them. Techniques include k-means clustering, support vector machines, principal components analysis and decision trees, for example. We will also look at applications of these techniques.
Course Resources
Special Educational Materials
UDOO Neo Ideal for IIoT Designs Where Extensive I/O Sensors and Powerful Compute Resources are KeyHow a New Breed of Wireless MCUs and Modules Are Enabling Connected IoT ApplicationsDevelop Powerful Event-Driven Video
September 15,
2016
Failure Analysis & Anomaly Detection
For this lecture we will talk about failure detection and anomaly detection, and the analyses used in conjunction with them. In the IoT we often want to perform these on a continuous basis. This introduces some interesting features in the analysis that we will discuss.
Course Resources
Special Educational Materials
UDOO Neo Ideal for IIoT Designs Where Extensive I/O Sensors and Powerful Compute Resources are KeyHow a New Breed of Wireless MCUs and Modules Are Enabling Connected IoT ApplicationsDevelop Powerful Event-Driven Video
September 16,
2016
Optimization
One major feature of the IoT is that we now have lots of data over long periods of time for many types of systems.We can use this rich data horde to perform optimizations in current systems and in the design of new or evolved systems.These techniques often use complex formulations involving linear and non-linear systems approaches.Fortunately, there are many existing implementations of these methods that make their execution relatively easy.We will look at them in a couple different languages.
Course Resources
Special Educational Materials
UDOO Neo Ideal for IIoT Designs Where Extensive I/O Sensors and Powerful Compute Resources are KeyHow a New Breed of Wireless MCUs and Modules Are Enabling Connected IoT ApplicationsDevelop Powerful Event-Driven Video
Instructor
Louis Giokas