C3_W1_Misunderstanding on how to explain the result

Course Q&A Machine Learning Specialization Unsupervised Learning, Recommenders, Reinforcement
1

Hi, I have a misunderstanding point. From course tutorial, it was mentioned that we can “find interesting structure from data”. After the algorithm generated the structure, how can we explain the result for publishing publications? Thanks.

2

Hello @kaian0414,

Show us more context. Which course video and at what time mark? Or which lab? Is there anything else you want to share that base on which you want us to discuss?

Raymond

3

hello @rmwkwok
Course 3 week 1, about the basic overview of unsupervised learning. Professor mentioned that we can find some structure from the data with unsuepervised learning but i have some confuse as mentioned in the question. i am not sure after we clustered the data into groups but how to explain the group scenarios? and how to explain for later publication. thanks

4

Hi @kaian0414,

I think you were talking about the C3 W1 Video “What is clustering?”, and from which specifically the following:

Because we don’t have target labels y, we’re not able to tell the algorithm what is the “right answer, y” that we wanted to predict. Instead, we’re going to ask the algorithm to find something interesting about the data, that is to find some interesting structure about this data. But the first unsupervised learning algorithm that you learn about is called a clustering algorithm, which looks for one particular type of structure in the data

Here the meaning of structure, I believe, is really how you may cluster the samples into different groups. The algorithms do not explain things, but based upon the grouping results, it is going to be your hard work to try to make up some explanations that are consistent with the principle of the algorithms. For example, if you have used K-means on a selected set of features, then your explanations are going to connect with the proximity among in-group samples over those features. You might make attempts to further narrow down the features so that your explanation may become more useful and actionable. You might support your explanation with the nature of the dataset and the features.

Therefore, although there is algorithm to cluster data points, there is no algorithm to produce that explanation and it will be your job.

Cheers,
Raymond

5

Oh I see. Thank you! I have such an annoying in mind for so long and you explained what my misunderstanding, haha, thanks again!!!