Course Description
This course covers data preparation for machine learning. The course includes example applications and example code. Best practices in telecommunication, data cleaning, data transformation, and how to turn the developed algorithms into a robust system are covered in depth. The benefits for taking this course include turning your users and system administrators into heroes, having a system that can remain viable, and successfully moving into the next great application while handing the system to lower level people.
What am I going to get from this course?
Select the appropriate robot for their application, integrate the real time data exchange at the appropriate frequency, include the predictors that may be needed for training, prepare the data for machine learning, and deploy robustly for success.
Prerequisites and Target Audience
What will students need to know or do before starting this course?
A knowledge of Python, C++, Kafka, et al, will help if the student really wants to be able to deploy, but these sections can be skipped for the functional user or executive who is interested at the survey level because the class is in the "for dummies" style.
Who should take this course? Who should not?
Because the course is in the "for dummies" style, anyone from functional staff to executives can take the course at the survey level. To really get to implement these topics, because the underlying math is not covered, a masters in quantitative methods is advised.
Curriculum
Module 1: Introduction
09:42
Lecture 1
What this class is
09:42
This lecture differentiates between programming robots versus using fully functioning robots for applications.
Module 2: Real time data exchange
01:11:19
Lecture 2
Robots with WiFi
06:28
Open source versus closed source robots with telecommunication is covered plus the current status of cellular on board.
Lecture 3
AI applications for robots are discussed.
17:14
A robot for machine learning as a class example is selected.
Finding robots with WiFi or cellular on board.
Lecture 5
Real Time Data Exchange
14:04
Robots with open source output files.
Lecture 6
Telecommunication
03:09
The current state of telecommunication on board robots.
Lecture 7
TC Do's and Don't's
02:28
Best practices to avoid pitfalls.
Selection of an example for class.
Lecture 9
Program platform
00:45
Lecture 10
Sample code
06:12
Lecture 11
Sample CSV file
14:09
Walkthrough of variables and format.
Lecture 12
Sample variable list
00:15
This list is for the class example application.
Module 3: Two examples
01:03:58
Lecture 13
Sample application 1
57:55
Frequency and variables for training the machine learning for application 1.
Lecture 14
Sample application 1
00:39
Frequency and variables for training the machine learning for application 2.
Lecture 15
Sample application 2
04:44
Submit the frequency and variables for training the machine learning for your application.
Module 4: Automation
22:50
Lecture 17
Data collection
00:19
Obtaining the samples for training and later for the automated system.
Lecture 18
Cloud storage
00:35
Best practices and revisions and upgrades.
Lecture 19
Pseudo real time collection
00:43
Real time to your application depends on the frequency of sampling.
Lecture 20
Exogenous factors
01:46
The environment defined by your selected predictors may widen over time.
Lecture 21
Data cleaning
01:06
Results are only as good as the inputs and your understanding of the application.
Lecture 22
Data transformations
15:43
Data usually needs to be transformed so that values which are too big or too small do not skew the results.
Lecture 23
Building the algorithm
00:06
The numerous tools available to understand the interactions of the predictors.
Lecture 24
Automation
00:20
How to construct the machine learning is covered in more depth in the next course the Industrial Internet of Things, AKA IIoT, AKA Machine to Machine, M2M.
Lecture 25
Machine learning
00:42
Neural networks are covered in more depth in the next class, IIoT.
Lecture 26
6 sigma alerts
00:31
SPC, AKA Statistical Process Control, is important for being able to manage when something in the predictors goes askew.
Logging sampling events is important when some predictor goes askew and needs investigation.
Lecture 28
Reporting
00:39
Being able to send logs and reports to the system user on demand or when a predictor goes askew via Android or iOS is key to keeping the system running 24/7.
Module 5: Deploying
02:50
Lecture 29
Documentation
00:31
Both user and system administrator documentation are important for successful deployment. This can be an AI project inside your machine learning application, LOL.
Lecture 30
Training
00:18
Best practices are suggested based on the number of locations in your installation and their geographic proximity to each other.
Lecture 31
Deploying
01:30
How to deploy and make your users and system administrators heroes, so that you can move onto your next exciting robot machine learning application.
Lecture 32
Updating
00:31
A system that is not updated is not being used and will die from disuse.