Will the course cover both supervised and unsupervised models?

Introduction: Machine Learning concepts (old)
#1

Great intro, but it made it sound like the course will focus mainly on supervised models.

Is this correct or I have I misunderstood?

#2

I heard the same, but it could be great that the course focuses on supervised and unsupervised machine learning

#3

This first MOOC is only focusing on supervised learning

3 Likes
#4

Is a second MOOC being planned already (which will cover unsupervised models and other topics)?

#5

We have in mind to work on an additional topic. For the moment this is unclear if it would be a separate MOOC or the current 2.0 of this MOOC. However, we did not discuss the topic to be included/added.

It should as well be noted that developing a MOOC is really time-consuming. The current MOOC is the result of a 1+ year of work, using some previous contributions.

1 Like
#6

I am surprised that supervised learning (SL) is being taught before unsupervised learning (UL). I say this because SL requires the target class, so during this entire MOOC, someone has to have created a target class for every exercise. But, does the real world always start with an available target class? If not, why start with SL? [Just terribly curious!]

#7

It is true that collecting data in real world usually does not provide the associated target and it is indeed one of the challenges of supervised learning.

However, in the real world, problems that we want to solve with data should come with a way to evaluate if our “solution” works: is our computed-aided diagnosis detecting cancers or is our spam system rejecting the right emails? To measure this metric, you need to measure a quantity between true and predicted targets, and so you need true target.

I am not sure that starting by unsupervised techniques will help at grasping easily real world problems that one tries to solve in data science: using k-means to cluster iris is it meaningful without stating anything about iris species? Does applying PCA would allow grasping anything regarding a real-world dataset?

Of course, unsupervised learning is part of the statistical learning toolbox and is useful but, in my humble opinion, I am unsure that one should start by teaching those techniques first.

1 Like
#8

Thanks for clarifying!

And thanks for the hard work!

#9

In truth, I don’t mind which you teach first, I am just very happy to be taking part in this MOOC. Thank you!

2 Likes