Week 3 "Decision boundary" video question

Course Q&A Machine Learning Specialization Supervised ML: Regression and Classification
1

So in week 3 in the video called “Decision boundary” there is a question:
“Let’s say you are creating a tumor detection algorithm. Your algorithm will be used to flag potential tumors for future inspection by a specialist. What value should you use for a threshold?”
The right answer is LOW(0.2), but shouldn’t it be HIGH(0.9)?
Because in previous videos we agreed that asking the question “If the tumor is malignant, if it’s yes, then y=1”. So doctor should notice malignant tumors first of all, isn’t it? Or I got some misunderstanding here?

2

Hello,

if tumor is malignant, it is yes then y=1, if no 0 that does not define the threshold.

threshold is the ability to find as small as possible defect, tumor, abnormality in a data, so as low as you can find, it is better.

please refer this Week 3 decision boundry sigmoid function - #2 by Deepti_Prasad

Hope you understood!!

Keep Learning!!!
Cheers
DP

3

I think I understand better now. So the point here is to catch anything higher than threshold, which is selected 0.2 in this case. So anything higher has a potential risk to be malignant, so needs doctors closer attention. But anything between 0 and 0.2 is considered low risk and can be skipped, right?

4

You got it right !!! In the question asked 0.2 as a threshold becomes the decision boundary for a tumor to be not missed out.

Although in scientific reasoning, 0.2 can be benign or malignant tumor based on the cell differentiation.

But in the question asked it is clearly deciding if threshold is 0.2 or higher, then y is 1, otherwise 0.

Regards
DP

1 Like