Dear Class,
Jon Reilly co-founder of NoCodeAI startup Akkio spoke to our class about AutoML today. This was a good next step to our past lessons on how to take the business problems for our Capstone Lab AI class and arrive at a Problem statement using the 4P Problem Statement Framework.
Jon Reilly reminded us about what is Machine Learning with a lens to look at the underlying training data that you will use to best fit a classification model using AutoML.
(Click on image below for Jon Reilly’s talk to the Capstone Lab class about AutoML/NoCodeAI)
Three key lessons about getting your data ready for Machine Learning are:
Check if any causal variables are included in the training data. This means that you add a column that overfits the data and allows some data to lead to an obvious prediction.
We had looking at an example of running a model on Akkio in the first class of Capstone Lab. Check it out.
With this we have wrapped the prep portion (or PowerPoint portion of building an AI as Jon Reilly put it) with learning to analyze a business problem, choosing multiple use cases, discussing business strategy on what is actionable with the AI’s prediction and arriving at a problem statement. We also have begun the data prep journey to get ready with training data.
Next step we jump into the hands-on lab with data visualization, data prep and wrangling, online lab setup and model fitting to complete our DataScience Sprint cycle to build an AI with a multi-disciplinary team for a real business problem.
Onward!